
Citation: Usman Mahmood, David W Johnson, Magid A Fahim. Cardiac biomarkers in dialysis[J]. AIMS Genetics, 2017, 4(1): 1-20. doi: 10.3934/genet.2017.1.1
[1] | Yan Ling Fu, Wei Zhang . Some results on frames by pre-frame operators in Q-Hilbert spaces. AIMS Mathematics, 2023, 8(12): 28878-28896. doi: 10.3934/math.20231480 |
[2] | Gang Wang . Some properties of weaving K-frames in n-Hilbert space. AIMS Mathematics, 2024, 9(9): 25438-25456. doi: 10.3934/math.20241242 |
[3] | Sergio Verdú . Relative information spectra with applications to statistical inference. AIMS Mathematics, 2024, 9(12): 35038-35090. doi: 10.3934/math.20241668 |
[4] | Ligong Wang . Output statistics, equivocation, and state masking. AIMS Mathematics, 2025, 10(6): 13151-13165. doi: 10.3934/math.2025590 |
[5] | Cure Arenas Jaffeth, Ferrer Sotelo Kandy, Ferrer Villar Osmin . Functions of bounded (2,k)-variation in 2-normed spaces. AIMS Mathematics, 2024, 9(9): 24166-24183. doi: 10.3934/math.20241175 |
[6] | Chibueze C. Okeke, Abubakar Adamu, Ratthaprom Promkam, Pongsakorn Sunthrayuth . Two-step inertial method for solving split common null point problem with multiple output sets in Hilbert spaces. AIMS Mathematics, 2023, 8(9): 20201-20222. doi: 10.3934/math.20231030 |
[7] | Osmin Ferrer Villar, Jesús Domínguez Acosta, Edilberto Arroyo Ortiz . Frames associated with an operator in spaces with an indefinite metric. AIMS Mathematics, 2023, 8(7): 15712-15722. doi: 10.3934/math.2023802 |
[8] | Abdullah Ali H. Ahmadini, Amal S. Hassan, Ahmed N. Zaky, Shokrya S. Alshqaq . Bayesian inference of dynamic cumulative residual entropy from Pareto Ⅱ distribution with application to COVID-19. AIMS Mathematics, 2021, 6(3): 2196-2216. doi: 10.3934/math.2021133 |
[9] | Messaoud Bounkhel . V-Moreau envelope of nonconvex functions on smooth Banach spaces. AIMS Mathematics, 2024, 9(10): 28589-28610. doi: 10.3934/math.20241387 |
[10] | Jamilu Adamu, Kanikar Muangchoo, Abbas Ja'afaru Badakaya, Jewaidu Rilwan . On pursuit-evasion differential game problem in a Hilbert space. AIMS Mathematics, 2020, 5(6): 7467-7479. doi: 10.3934/math.2020478 |
In [13,14,15,16,19], it was proposed that insight into a probability distribution, μ, posed on a Hilbert space, H, could be obtained by finding a best fit Gaussian approximation, ν. This notion of best, or optimal, was with respect to the relative entropy, or Kullback-Leibler divergence:
R(ν||μ)={Eν[logdνdμ],ν≪μ,+∞,otherwise. | (1.1) |
Having a Gaussian approximation provides qualitative insight into μ, as it provides a concrete notion of the mean and variance of the distribution. Additionally, this optimized distribution can be used in algorithms, such as random walk Metropolis, as a preconditioned proposal distribution to improve performance. Such a strategy can benefit a number of applications, including path space sampling for molecular dynamics and parameter estimation in statistical inverse problems.
Observe that in the definition of R, (1.1), there is an asymmetry in the arguments. Were we to work with R(μ||ν), our optimal Gaussian would capture the first and second moments of μ, and in some applications this is desirable. However, for a multimodal problem (consider a distribution with two well separated modes), this would be inadequate; our form attempts to match individual modes of the distribution by a Gaussian. For a recent review of the R(ν||μ) problem, see [4], where it is remarked that this choice of arguments is likely to underestimate the dispersion of the distribution of interest, μ. The other ordering of arguments has been explored, in the finite dimensional case, in [2,3,10,18].
To be of computational use, it is necessary to have an algorithm that will converge to this optimal distribution. In [15], this was accomplished by first expressing ν=N(m,C(p)), where m is the mean and p is a parameter inducing a well defined covariance operator, and then solving the problem,
(m,p)∈argminR(N(m,C(p))||μ), | (1.2) |
over an admissible set. The optimization step itself was done using the Robbins-Monro algorithm (RM), [17], by seeking a root of the first variation of the relative entropy. While the numerical results of [15] were satisfactory, being consistent with theoretical expectations, no rigorous justification for the application of RM to the examples was given.
In this work, we emphasize the study and application of RM to potentially infinite dimensional problems. Indeed, following the framework of [15,16], we assume that μ is posed on the Borel σ-algebra of a separable Hilbert space (H,⟨∙,∙⟩,‖∙‖). For simplicity, we will leave the covariance operator C fixed, and only optimize over the mean, m. Even in this case, we are seeking m∈H, a potentially infinite-dimensional space.
Given the objective function f:H→H, assume that it has a root, x⋆. In our application to relative entropy, f will be its first variation. Further, we assume that we can only observe a noisy version of f, F:H×χ→H, such that for all x∈H,
f(x)=E[F(x,Z)]=∫χF(x,z)μZ(dz), | (1.3) |
where μZ is the distribution associated with the random variable (r.v.) Z, taking values in the auxiliary space χ. The naive Robbins-Monro algorithm is given by
Xn+1=Xn−an+1F(Xn,Zn+1), | (1.4) |
where Zn∼μZ, are independent and identically distributed (i.i.d.), and an>0 is a carefully chosen sequence. Subject to assumptions on f, F, and the distribution μZ, it is known that Xn will converge to x⋆ almost surely (a.s.), in finite dimensions, [5,6,17]. Often, one needs to assume that f grows at most linearly,
‖f(x)‖≤c0+c1‖x‖, | (1.5) |
in order to apply the results in the aforementioned papers. The analysis in the finite dimensional case has been refined tremendously over the years, including an analysis based on continuous dynamical systems. We refer the reader to the books [1,8,11] and references therein.
As noted, much of the analysis requires the regression function f to have, at most, linear growth. Alternatively, an a priori assumption is sometimes made that the entire sequence generated by (1.4) stays in a bounded set. Both assumptions are limiting, though, in practice, one may find that the algorithms converge.
One way of overcoming these assumptions, while still ensuring convergence, is to introduce trust regions that the sequence {Xn} is permitted to explore, along with a "truncation" which enforces the constraint. Such truncations distort (1.4) into
Xn+1=Xn−an+1F(Xn,Zn+1)+an+1Pn+1, | (1.6) |
where Pn+1 is the projection keeping the sequence {Xn} within the trust region. Projection algorithms are also discussed in [1,8,11].
We consider RM on a possibly infinite dimensional separable Hilbert space. This is of particular interest as, in the context of relative entropy optimization, we may be seeking a distribution in a Sobolev space associated with a PDE model. A general analysis of RM with truncations in Hilbert spaces can be found in [20]. The main purpose of this work is to adapt the analysis of [12] to the Hilbert space setting for two versions of the truncated problem. The motivation for this is that the analysis of [12] is quite straightforward, and it is instructive to see how it can be easily adapted to the infinite dimensional setting. The key modification in the proof is that results for Banach space valued martingales must be invoked. We also adapt the results to a version of the algorithm where there is prior knowledge on the location of the root. With these results in hand, we can then verify that the relative entropy minimization problem can be solved using RM.
In some problems, one may have a priori information on the root. For instance, we may know that x⋆∈U1, some open bounded set. In this version of the truncated algorithm, we have two open bounded sets, U0⊊U1, and x⋆∈U1. Let σ0=0 and X0∈U0 be given, then (1.6) can be formulated as
˜Xn+1=Xn−an+1F(Xn,Zn+1) | (1.7a) |
Xn+1={˜Xn+1˜Xn+1∈U1X(σn)0˜Xn+1∉U1 | (1.7b) |
σn+1={σn˜Xn+1∈U1σn+1˜Xn+1∉U1 | (1.7c) |
We interpret ˜Xn+1 as the proposed move, which is either accepted or rejected depending on whether or not it will remain in the trust region. If it is rejected, the algorithm restarts at X(σn)0∈U0. The restart points, {X(σn)0}, may be random, or it may be that X(σn)0=X0 is fixed. The essential property is that the algorithm will restart in the interior of the trust region, away from its boundary. The r.v. σn counts the number of times a truncation has occurred. Algorithm (1.7) can now be expressed as
Xn+1=Xn−an+1F(Xn,Zn+1)+Pn+1Pn+1={X(σn)0−˜Xn+1}1˜Xn+1∉U1. | (1.8) |
In the second version of truncated Robbins-Monro, define the sequence of open bounded sets, Un such that:
U0⊊U1⊊U2⊊…,∪∞n=0Un=H. | (1.9) |
Again, letting X0∈U0, σ0=0, the algorithm is
˜Xn+1=Xn−an+1F(Xn,Zn+1) | (1.10a) |
Xn+1={˜Xn+1˜Xn+1∈UσnX(σn)0˜Xn+1∉Uσn | (1.10b) |
σn+1={σn˜Xn+1∈Uσnσn+1˜Xn+1∉Uσn | (1.10c) |
A consequence of this formulation is that Xn∈Uσn for all n. As before, the restart points may be random or fixed, and they are in U0. This would appear superior to the fixed trust region algorithm, as it does not require knowledge of the sets. However, to guarantee convergence, global (in H) assumptions on the regression function are required; see Assumption 2 below. (1.10) can written with Pn+1 as
Xn+1=Xn−an+1F(Xn,Zn+1)+Pn+1Pn+1={X(σn)0−˜Xn+1}1˜Xn+1∉Uσn | (1.11) |
In Section 2, we state sufficient assumptions for which we are able to prove convergence in both the fixed and expanding trust region problems, and we also establish some preliminary results. In Section 3, we focus on the relative entropy minimization problem, and identify what assumptions must hold for convergence to be guaranteed. Examples are then presented in Section 4, and we conclude with remarks in Section 5.
We first reformulate (1.8) and (1.15) in the more general form
Xn+1=Xn−an+1f(Xn)−an+1δMn+1⏟=˜Xn+1+an+1Pn+1, | (2.1) |
where δMn+1, the noise term, is
δMn+1=F(Xn,Zn+1)−f(Xn)=F(Xn,Zn+1)−E[F(Xn,Zn+1)∣Xn]. | (2.2) |
A natural filtration for this problem is Fn=σ(X0,Z1,…,Zn). Xn is Fn measurable and the noise term can be expressed in terms of the filtration as δMn+1=F(Xn,Zn+1)−E[F(Xn,Zn+1)∣Fn].
We now state our main assumptions:
Assumption 1. f has a zero, x⋆. In the case of the fixed trust region problem, there exist R0<R1 such that
U0⊆BR0(x⋆)⊂BR1(x⋆)⊆U1. |
In the case of the expanding trust region problem, the open sets are defined as Un=Brn(0) with
0<r0<r1<r2<…<rn→∞. | (2.3) |
These sets clearly satisfy (1.9).
Assumption 2. For any 0<a<A, there exists δ>0:
inf |
In the case of the fixed truncation, this inequality is restricted to x\in U_1 . This is akin to a convexity condition on a functional \mathcal{F} with f = D\mathcal{F} .
Assumption 3. x\mapsto \mathbb{E}[\left \|{F(x, Z)}\right\|^2] is bounded on bounded sets, with the restriction to U_1 in the case of fixed trust regions.
Assumption 4. a_n > 0 , \sum a_n = \infty , and \sum a_n^2 < \infty
Theorem 2.1. Under the above assumptions, for the fixed trust region problem, X_n \to x_\star a.s. and \sigma_n is a.s. finite.
Theorem 2.2. Under the above assumptions, for the expanding trust region problem, X_n \to x_\star a.s. and \sigma_n is a.s. finite.
Note the distinction between the assumptions in the two algorithms. In the fixed truncation algorithm, Assumptions 2 and 3 need only hold in the set U_1 , while in the expanding truncation algorithm, they must hold in all of \mathcal{H} . While this would seem to be a weaker condition, it requires identification of the sets U_0 and U_1 for which the assumptions hold. Such sets may not be readily identifiable, as we will see in our examples.
We first need some additional information about f and the noise sequence \delta M_n .
Lemma 2.1. Under Assumption 3, f is bounded on U_1 , for the fixed trust region problem, and on arbitrary bounded sets, for the expanding trust region problem.
Proof. Trivially,
\begin{equation*} \left \|{f(x)}\right\| = \left \|{ \mathbb{E}[F(x,Z)]}\right\|\leq \mathbb{E}[\left \|{F(x,Z)}\right\| ]\leq \sqrt{ \mathbb{E}[\left \|{F(x,Z)}\right\|^2}], \end{equation*} |
and the results follows from the assumption.
Proposition 2.1. For the fixed trust region problem, let
M_n = \sum\limits_{i = 1}^n a_i \delta M_i. |
Alternatively, in the expanding trust region problem, for r > 0 , let
M_n = \sum\limits_{i = 1}^n a_i \delta M_i 1_{\left \|{X_{i-1}-x_\star}\right\|\leq r}. |
Under Assumptions 3 and 4, M_n is a martingale, converging in \mathcal{H} , a.s.
Proof. The following argument holds in both the fixed and expanding trust region problems, with appropriate modifications. We present the expanding trust region case. The proof is broken up into 3 steps:
1. Relying on Theorem 6 of [7] for Banach space valued martingales, it will be sufficient to show that M_n is a martingale, uniformly bounded in L^1(\mathbb{P}) .
2. In the case of the expanding truncations,
\begin{equation*} \begin{split} \mathbb{E}[\left \|{\delta M_i 1_{\left\| {{X_{i - 1}} - {x_\star}} \right\| \le r}}\right\|^2] &\leq 2 \mathbb{E}[\left \|{F(X_{i-1},Z_i)1_{\left\| {{X_{i - 1}} - {x_\star}} \right\| \le r}}\right\|^2] + 2 \mathbb{E}[\left \|{f(X_{i-1})1_{\left\| {{X_{i - 1}} - {x_\star}} \right\| \le r}}\right\|^2] \\ &\leq 2\sup\limits_{\left \|{x- x_\star}\right\|\leq r} \mathbb{E}[\left \|{F(x,Z)}\right\|^2] + 2\sup\limits_{\left \|{x- x_\star}\right\|\leq r}\left \|{f(x)}\right\|^2 \end{split} \end{equation*} |
Since both of these terms are bounded, independently of i , by Assumption 3 and Lemma 1, this is finite.
3. Next, since \{\delta M_i 1_{\|X_{i-1}-x_\star\|\leq r}\} is a martingale difference sequence, we can use the above estimate to obtain the uniform L^2(\mathbb{P}) bound,
\begin{equation*} \begin{split} \mathbb{E}[\left \|{M_n}\right\|^2]& = \sum\limits_{i = 1}^na_i^2 \mathbb{E}[\left \|{{\delta M_i 1_{\left\| {{X_{i - 1}} - {x_\star}} \right\| \le r}}}\right\|^2] \leq\sup\limits_{i} \mathbb{E}[\left \|{{\delta M_i 1_{\left\| {{X_{i - 1}} - {x_\star}} \right\| \le r}}}\right\|^2]\sum\limits_{i = 1}^\infty a_i^2 \lt \infty \end{split} \end{equation*} |
Uniform boundedness in L^2 , gives boundedness in L^1 , and this implies a.s. convergence in \mathcal{H} .
In this section we prove results showing that only finitely many truncations will occur, in either the fixed or expanding trust region case. Recall that when a truncation occurs, the equivalent conditions hold: P_{n+1}\neq 0 ; \sigma_{n+1} = \sigma_n+1 ; and \tilde{X}_{n+1}\notin U_1 in the fixed trust region algorithm, while \tilde{X}_{n+1}\notin U_{\sigma_n} in the expanding trust region case.
Lemma 2.2. In the fixed trust region algorithm, if Assumptions 1, 2, 3, and 4 hold, then the number of truncations is a.s. finite; a.s., there exists N , such that for all n\geq N , \sigma_n = \sigma_N .
Proof. We break the proof up into 7 steps:
1. Pick \rho and \rho' such that
\begin{equation} R_0 \lt \rho' \lt \rho \lt R_1 \end{equation} | (2.4) |
Let \bar f = \sup\|f(x)\| , with the supremum over U_1 ; this bound exists by Lemma 1. Under Assumption 2, there exists \delta > 0 such that
\begin{equation} \inf\limits_{R_0/2 \leq \|x-x_\star\|\leq R_1}\left\langle {x-x_\star},{f(x)}\right\rangle = \delta. \end{equation} | (2.5) |
Having fixed \rho , \rho' , \bar f , and \delta , take \epsilon > 0 such that:
\begin{equation} \epsilon \lt \min\left\{{\rho'-R_0, \frac{R_1-\rho'}{2 + \bar f}, \frac{\rho' - R_0}{\bar f}, \frac{R_0}{2}, \frac{\delta}{2\bar{f}}, \frac{\delta}{\bar{f}^2}, {\rho - \rho'}}\right\}. \end{equation} | (2.6) |
Having fixed such an \epsilon , by the assumptions of this lemma and Proposition 1, a.s., there exists n_{ \epsilon} such that for any n, m\geq n_{ \epsilon} , both
\begin{equation} \left \|{\sum\limits_{k = n}^m a_k \delta M_k}\right\|\leq \epsilon, \quad a_n \leq \epsilon. \end{equation} | (2.7) |
2. Define the auxiliary sequence
\begin{equation} X_n' = X_n - \sum\limits_{k = n+1}^\infty a_k \delta M_k. \end{equation} | (2.8) |
Using (2.1), we can then write
\begin{equation} X_{n+1}' = X_{n}' - a_{n+1}f(X_n) +a_{n+1}P_{n+1}. \end{equation} | (2.9) |
By (2.7), for any n \geq n_ \epsilon ,
\begin{equation} \|X_n' - X_n\|\leq \epsilon \end{equation} | (2.10) |
3. We will show X_n' \in B_{\rho'}(x_\star) for all n large enough. The significance of this is that if n\geq n_ \epsilon , and X_n'\in B_{\rho'}(x_\star) , then no truncation occurs. Indeed, using (2.6)
\begin{equation} \begin{split} \|\tilde{X}_{n+1} -x_\star\| &\leq \|X_{n}' -x_\star\| +\|X_n - X_n'\| + a_{n+1}\bar f + \|a_{n+1}\delta M_{n+1}\|\\ & \lt \rho' + \epsilon + \epsilon \bar f + \epsilon \lt R_1, \Rightarrow \tilde{X}_{n+1}\in U_1. \end{split} \end{equation} | (2.11) |
Consequently, P_{n+1} = 0 , X_{n+1} = \tilde{X}_{n+1} , and \sigma_{n+1} = \sigma_n . Thus, establishing X_n'\in B_{\rho'}(x_\star) will yield the result.
4. Let
\begin{equation} N = \inf\{n\geq n_ \epsilon\mid \tilde{X}_{n+1}\not\in U_1\}+1 \end{equation} | (2.12) |
This corresponds to the the first truncation after n_ \epsilon . If the above set is empty, for that realization, no truncations occur after n_ \epsilon , and we are done. In such a case, we may take N = n_ \epsilon in the statement of the lemma.
5. We now prove by induction that in the case that (2.12) is finite, X_n'\in B_{\rho'}(x_\star) for all n\geq N . First, note that X_N \in B_{R_0}(x_\star)\subset B_{\rho}(x_\star) . By (2.6) and (2.10),
\|X_{N}'-x_\star\| \leq \|X_{N}-x_\star\| + \|X_{N}'- X_{N}\| \lt R_0 + \epsilon \lt \rho', \Rightarrow X_N' \in B_{\rho'}(x_\star). |
Next, assume X_{N}', X_{N+1}', \ldots, X_n' are all in B_{\rho'}(x_\star) . Using (2.11), we have that P_{N+1} = \ldots = P_{n+1} = 0 and \sigma_{N} = \ldots = \sigma_{n} = \sigma_{n+1} . Therefore,
\begin{equation} \begin{split} \left \|{X_{n+1}' - x_\star}\right\|^2 & = \left \|{X'_n - x_\star}\right\|^2 - 2a_{n+1} \left\langle {X'_{n}-x_\star},{f(X_n)}\right\rangle + a_{n+1}^2 \left \|{f(X_n)}\right\|^2\\ &\leq \left \|{X'_n - x_\star}\right\|^2 - 2a_{n+1} \left\langle {X'_{n}-x_\star},{f(X_n)}\right\rangle + a_{n+1} \epsilon \bar f^2 \end{split} \end{equation} | (2.13) |
We now consider two cases of (2.13) to conclude \|X_{n+1}' - x_\star\| < \rho' .
6. In the first case, \|X'_n - x_\star\|\leq R_0 . By Cauchy-Schwarz and (2.6)
\|X_{n+1}' - x_\star\|^2 \lt R_0^2 + 2 \epsilon R_0 \bar f + \epsilon^2\bar{f}^2 = (R_0 + \epsilon \bar f)^2 \lt (\rho')^2. |
In the second case, R_0 < \|X'_n - x_\star\| < \rho' . Dissecting the inner product term in (2.13) and using Assumption 2 and (2.10),
\begin{equation} \begin{split} \left\langle {X'_{n}-x_\star},{f(X_n) }\right\rangle & = \left\langle {X_{n}-x_\star},{f(X_n) }\right\rangle + \left\langle {X'_{n}-X_n},{f(X_n) }\right\rangle\\ &\geq\left\langle {X_{n}-x_\star},{f(X_n) }\right\rangle-\bar f \epsilon \end{split} \end{equation} | (2.14) |
Conditions (2.6) and (2.10) yield the following upper and lower bounds:
\begin{align*} \|X_n-x_\star\|&\geq\|X_n'-x_\star\| -\|X_n'-X_n\| \geq R_0 - \epsilon \gt \tfrac{1}{2}R_0, \\ \|X_n-x_\star\|&\leq \|X_n'-x_\star\| +\|X_n'-X_n\| \leq \rho' + \epsilon \lt \rho \lt R_1. \end{align*} |
Therefore, (2.5) applies and \left\langle {X_{n}-x_\star}, {f(X_n)}\right\rangle\geq \delta . Using this in (2.14), and condition (2.6),
\begin{equation*} \left\langle {X'_{n}-x_\star},{f(X_n) }\right\rangle\geq \delta - \bar f \epsilon \gt \tfrac{1}{2}\delta. \end{equation*} |
Substituting this last estimate back into (2.13), and using (2.6),
\|X_{n+1}' - x_\star\|^2 \lt (\rho')^2 - a_{n+1}(\delta - \epsilon\bar f^2) \lt (\rho')^2. |
This completes the inductive step.
7. Since the auxiliary sequence remains in B_{\rho'}(x_\star) for all n\geq N > n_ \epsilon , (2.11) ensures \tilde X_{n+1}\in B_{R_1}(x_\star) , P_{n+1} = 0 , and \sigma_{n+1} = \sigma_N , a.s.
To obtain a similar result for the expanding trust region problem, we first relate the finiteness of the number of truncations with the sequence persisting in a bounded set.
Lemma 2.3. In the expanding trust region algorithm, if Assumptions 1, 3, and 4 hold, then the sequence remains in a set of the form B_{R}(0) for some R > 0 if and only if the number of truncations is finite, a.s.
Proof. We break this proof into 4 steps:
1. If the number of truncations is finite, then there exists N such that for all n\geq N , \sigma_n = \sigma_{N} . Consequently, the proposed moves are always accepted, and X_{n} \in U_{\sigma_{n}} = U_{\sigma_N} for all n\geq N . Since X_n \in U_{\sigma_n}\subset U_{\sigma_N} for n < N , X_n \in U_{\sigma_N} for all n . By Assumption 3, B_R(0) = B_{r_{\sigma_N}}(0) = U_{\sigma_N} is the desired set.
2. For the other direction, assume that there exists R > 0 such that X_n \in B_R(0) for all n . Since the r_n in (2.3) tend to infinity, there exists N_1 , such that R < R +1 < r_{N_1} . Hence, for all n\geq N_1 ,
\begin{equation} B_{R}(0)\subset B_{R+1}(0) \subset U_{n} \end{equation} | (2.15) |
Let \bar f = \sup \|f(x)\| , with the supremum over B_{R}(0) . Let \tilde{R} be sufficiently large such that B_{R+1}(0)\subset B_{\tilde R}(x_\star) . Lastly, using Proposition 1 and Assumption 4, a.s., there exists N_2 , such that for all n\geq N_2
\begin{equation} \| a_{n} \delta M_{n}1_{\|X_n - x_\star\|\leq \tilde R}\| \lt \frac{1}{2}, \quad a_{n} \lt \frac{1}{2(1+ \bar f)} \end{equation} | (2.16) |
Since X_n \in B_{R}(0) \subset B_{\tilde R}(x_\star) , the indicator function in (2.16) is always one, and \| a_{n} \delta M_{n}\| < 1/2 .
3. Next, let
\begin{equation} N = \inf \{ n\geq 0\mid \sigma_n\geq \max\{N_1, N_2\}\} \end{equation} | (2.17) |
If the above set is empty, then \sigma_n < \max\{N_1, N_2\} for all n , and the number of truncations is a.s. finite. In this case, the proof is complete.
4. If the set in (2.17) is not empy, then N < \infty . Take n\geq N . As X_n \in B_R(0) , and since n\geq \sigma_n\geq \max\{N_1, N_2\} , (2.16) applies. Therefore,
\begin{equation} \begin{split} \|\tilde{X}_{n+1}\| &\leq \|X_n\| + \| \tilde{X}_{n+1} - X_n\|\\ &\leq \|X_n\| + a_{n+1}\|f(X_n)\| + \| a_{n+1} \delta M_{n+1}\|\\ & \lt R + \tfrac{1}{2}+\tfrac{1}{2} \lt R +1. \end{split} \end{equation} | (2.18) |
Thus, \tilde{X}_{n+1}\in B_{R+1}(0)\subset U_{N_1} , \sigma_n \geq N_1 , and U_{N_1}\subset U_{\sigma_n} . Therefore, \tilde{X}_{n+1}\in U_{\sigma_n} . No truncation occurs, and \sigma_n = \sigma_{n+1} . Since this holds for all n\geq N , \sigma_n = \sigma_N , and the number of truncations is a.s. finite.
Next, we establish that, subject to an additional assumption, the sequence remains in a bounded set; the finiteness of the truncations is then a corollary.
Lemma 2.4. In the expanding trust region algorithm, if Assumptions 1, 2, 3, and 4 hold, and for any r > 0 , there a.s. exists N < \infty , such that for all n \geq N ,
\begin{equation*} P_{n+1} 1_{\|X_n- x_\star\|\leq r} = 0, \end{equation*} |
then \{X_n\} remains in a bounded open set, a.s.
Proof. We break this proof into 7 steps:
1. We begin by setting some constants for the rest of the proof. Fix R > 0 sufficiently large such that B_{R}(x_\star)\supset U_0 . Next, let \bar f = \sup {\|f(x)\|} with the supremum taken over B_{R+2}(x_\star) . Assumption 2 ensures there exists \delta > 0 such that
\begin{equation} \inf\limits_{R/2 \leq \|x-x_\star\|\leq R+2}\left\langle {x-x_\star},{f(x)}\right\rangle = \delta. \end{equation} | (2.19) |
Having fixed R , \bar f , and \delta , take \epsilon > 0 such that:
\begin{equation} \epsilon \lt \min\left\{{1,\frac{1}{\bar{f}}, \frac{\delta}{2\bar{f}}, \frac{\delta}{\bar{f}^2}, \frac{R}{2}}\right\}. \end{equation} | (2.20) |
By the assumptions of this lemma and Proposition 1 there exists, a.s., n_ \epsilon\geq N such that for all n\geq n_ \epsilon ,
\begin{gather} \left \|{\sum\limits_{i = n+1}^\infty a_i \delta M_i 1_{\left\| {{X_{i - 1}} - {x_\star}} \right\| \le R + 2}}\right\|\leq \epsilon, \end{gather} | (2.21a) |
\begin{gather} P_{n+1} 1_{\left \|{X_{n}- x_\star}\right\|\leq R+2} = 0, \end{gather} | (2.21b) |
\begin{gather} a_{n+1}\leq \epsilon \end{gather} | (2.21c) |
2. Define the modified sequence for n \geq n_ \epsilon as
\begin{equation} X_n' = X_n - \sum\limits_{k = n+1}^\infty a_k \delta M_k 1_{\left \|{X_{k-1} - x_\star}\right\|\leq R+2}, \Rightarrow \|X_n' - X_n\|\leq \epsilon. \end{equation} | (2.22) |
Using (2.1), we have the iteration
\begin{equation} X_{n+1}' = X_n' - a_{n+1} \delta M_{n+1} 1_{\left \|{X_n - x_\star}\right\| \gt R+2} - a_{n+1}f(X_n) +a_{n+1} P_{n+1}. \end{equation} | (2.23) |
3. Let
\begin{equation} N = \inf\{n\geq n_ \epsilon\mid \sigma_{n+1}\neq \sigma_n \}+1, \end{equation} | (2.24) |
the first time after n_ \epsilon that a truncation occurs.
If the above set is empty, no truncations occur after n_ \epsilon . In this case, \sigma_n = \sigma_{n_ \epsilon}\leq n_ \epsilon < \infty for all n \geq n_ \epsilon . Therefore, for all n\geq n_ \epsilon , X_n \in U_{\sigma_n}\subset U_{\sigma_{n_ \epsilon}} . Since U_{\sigma_n}\subset U_{\sigma_{n_ \epsilon}} for all n < n_ \epsilon too, the proof is complete in this case.
4. Now assume that N < \infty . We will show that \{X_n'\} remains in B_{R+1}(x_\star) for all n\geq N . Were this to hold, then for n\geq N ,
\begin{equation} \begin{split} \|X_n - x_\star\| &\leq \|X_n' - x_\star\| + \left \|{\sum\limits_{i = n+1}^\infty a_i \delta M_i 1_{\left\| {{X_{i - 1}} - {x_\star}} \right\| \le R + 2}}\right\|\\ & \lt R+1 + \epsilon \lt R+2, \end{split} \end{equation} | (2.25) |
having used (2.21) and (2.22). For n < N , X_n \in U_{\sigma_n}\subset U_{\sigma_N} = B_{r_N}(0) . Therefore, for all n , X_n \in B_{\tilde{R}}(0) where \tilde{R} = \max\{ r_N, \|x_\star\| + R+2\} .
5. We prove X_n' \in B_{R+1}(x_\star) by induction. First, since \epsilon < 1 and X_N \in U_{0}\subset B_R(x_\star) ,
\begin{equation*} \|X_N'-x_\star\|\leq \|X_N'-X_N\|+ \|X_N-x_\star\| \lt \epsilon + R \lt R+1. \end{equation*} |
Next, assume that X_N', X_{N+1}', \ldots, X_n' are all in B_{R+1}(x_\star) . By (2.25), X_n\in B_{R+2}(x_\star) . Since P_{n+1} 1_{\left \|{X_{n}- x_\star}\right\|\leq R+2} = 0 , we conclude P_{n+1} = 0 . The modified iteration (2.23) simplifies to have
X_{n+1}' = X_n' -a_{n+1}f(X_n), |
and
\begin{equation} \begin{split} \|X_{n+1}' - x_\star\|^2 & = \|X'_n - x_\star\|^2 - 2a_{n+1} \left\langle {X'_{n}-x_\star},{f(X_n) }\right\rangle\\ &\quad + a_{n+1}^2 \|f(X_n)\|^2\\ & \lt \|X'_n - x_\star\|^2 - 2a_{n+1} \left\langle {X'_{n}-x_\star},{f(X_n) }\right\rangle\\ &\quad + a_{n+1} \epsilon \bar{f}^2. \end{split} \end{equation} | (2.26) |
6. We now consider two cases of (2.26). First, assume \|X'_{n} - x_\star\|\leq R . Then (2.26) can immediately be bounded as
\begin{equation*} \begin{split} \|X_{n+1}' - x_\star\|^2& \lt R^2 + 2 \epsilon R \bar{f} + \epsilon^2\bar{f}^2 = (R+ \epsilon \bar{f})^2 \lt (R+1)^2, \end{split} \end{equation*} |
where we have used condition (2.20) in the last inequality.
7. Now consider the case R < \|X'_{n} - x_\star\| < R+1 . Using (2.20), the inner product in (2.26) can first be bounded from below:
\begin{equation*} \begin{split} \left\langle {X'_{n}-x_\star},{f(X_n) }\right\rangle & = \left\langle {X_{n}-x_\star},{f(X_n) }\right\rangle + \left\langle {X'_{n}-X_n},{f(X_n) }\right\rangle\\ &\geq \left\langle {X_{n}-x_\star},{f(X_n) }\right\rangle - \epsilon \bar{f} \gt \left\langle {X_{n}-x_\star},{f(X_n) }\right\rangle - \tfrac{1}{2}\delta. \end{split} \end{equation*} |
Next, using (2.20)
\begin{equation*} \|X_{n} - x_\star\|\geq \|X_{n}'-x_\star\| - \|X_{n} -X'_{n}\| \gt R- \epsilon \gt R - \tfrac{1}{2}R = \tfrac{1}{2}R \end{equation*} |
Therefore, \tfrac{1}{2}R < \|X_n -x_\star\| < R+2 , so (2.19) ensures \left\langle {X_{n}-x_\star}, {f(X_n)}\right\rangle\geq\delta and
\begin{equation*} \left\langle {X'_{n}-x_\star},{f(X_n) }\right\rangle \gt \delta-\tfrac{1}{2}\delta = \tfrac{1}{2}\delta. \end{equation*} |
Returning to (2.26), by (2.20),
\begin{equation*} \|X_{n+1}' - x_\star\|^2\leq (R+1)^2 - a_{n+1}(\delta - \epsilon \bar{f}^2) \lt (R+1)^2. \end{equation*} |
This completes the proof of the inductive step in this second case, completing the proof.
Corollary 2.1. For the expanding trust region algorithm, if Assumptions 1, 2, 3, and 4 hold, then the number of truncations is a.s. finite.
Proof. The proof is by contradiction. We break the proof into 4 steps:
1. Assuming that there are infinitely many truncations, Lemma 3 implies that the sequence cannot remain in a bounded set. Then, continuing to assume that Assumptions 1, 2, 3, and 4 hold, the only way for the conclusion of Lemma 4 to fail is if the assumption on P_{n+1} 1_{\|X_n - x_\star\|\leq r } is false. Therefore, there exists r > 0 and a set of positive measure on which a subsequence, P_{n_k+1} 1_{\|X_{n_k} - x_\star\|\leq r }\neq 0 . Hence X_{n_k}\in B_{r}(x_\star) , and P_{n_k+1}\neq 0 . So truncations occur at these indices, and \tilde{X}_{n_k+1} \not\in U_{\sigma_{n_k}} .
2. Let \bar{f} = \sup \|f(x)\| with the supremum over the set B_r(x_\star) and let \epsilon > 0 satisfy
\begin{equation} \epsilon \lt (\bar{f}+1)^{-1}. \end{equation} | (2.27) |
By our assumptions of the lemma and Proposition 1, there exists n_ \epsilon such that for all n\geq n_ \epsilon
\begin{equation} \|a_{n+1}\delta M_{n+1}1_{\|X_n - x_\star\|\leq r}\|\leq \epsilon, \quad a_{n+1}\leq \epsilon \end{equation} | (2.28) |
Along the subsequence, for all n_k\geq n_ \epsilon ,
\begin{equation} \|a_{n_k+1}\delta M_{n_k+1}1_{\|X_{n_k} - x_\star\|\leq r}\| = \|a_{n_k+1}\delta M_{n_k+1}\|\leq \epsilon. \end{equation} | (2.29) |
3. Furthermore, for n_k \geq n_ \epsilon :
\begin{equation} \begin{split} \|\tilde{X}_{n_k+1} - x_\star\| &\leq \| X_{n_k} - x_\star\| + a_{n_k+1}\|f(X_{n_k})\| + \|a_{n_k+1}\delta M_{n_k+1}\|\\ & \lt r + \epsilon \bar{f} + \epsilon \lt r+1, \Rightarrow \tilde{X}_{n_k+1} \in B_{r+1}(x_\star), \end{split} \end{equation} | (2.30) |
where (2.27) has been used in the last inequality.
4. By the definition of the U_n , there exists an index M such that U_{M}\supset B_{r+1}(x_\star) . Let
\begin{equation} N = \inf\{n\geq n_ \epsilon\mid \sigma_n \geq M\}. \end{equation} | (2.31) |
This set is nonempty and N < \infty since we have assumed there are infinitely many truncations. Let n_k \geq N . Then \sigma_{n_k}\geq M and U_{\sigma_{n_k}}\supset B_{r+1}(x_\star) . But (2.30) then implies that \tilde{X}_{n_k+1} \in U_{\sigma_{n_k}} , and no truncation will occur; P_{n_k+1} = 0 , providing the desired the contradiction.
Using the above results, we are able to prove Theorems 2.1 and 2.2. Since the proofs are quite similar, we present the more complicated expanding trust region case.
Proof. We split this proof into 6 steps:
1. First, by Corollary 1, only finitely many truncations occur. By Lemma 3, there exists R > 0 such that X_n\in B_R(0) for all n . Consequently, there is an r such that X_n\in B_r(x_\star) for all n .
2. Next, we fix constants. Let \bar f = \sup \|f(x)\| with the supremum taken over B_r(x_\star) . Fix \eta \in (0, 2R) , and use Assumption 2 to determine \delta > 0 such that
\begin{equation} \inf\limits_{\eta/2 \leq \|x-x_\star\|\leq r}\left\langle {x-x_\star},{f(x)}\right\rangle = \delta \end{equation} | (2.32) |
Take \epsilon > 0 such that:
\begin{equation} \epsilon \lt \min\left\{{1,\frac{\eta}{2}, \frac{\delta}{2\bar{f}}, \frac{\delta}{2\bar{f}^2}}\right\} \end{equation} | (2.33) |
Having set \epsilon , we again appeal to Assumption 4 and Proposition 1 to find n_ \epsilon such that for all n\geq n_ \epsilon :
\begin{equation} \left \|{\sum\limits_{i = n+1}^\infty a_i \delta M_i 1_{\|X_{i-1} -x_\star\|\leq r}}\right\| = \left \|{\sum\limits_{i = n+1}^\infty a_i \delta M_i }\right\| \leq \epsilon, \quad a_{n+1}\leq \epsilon \end{equation} | (2.34) |
3. Define the auxiliary sequence,
\begin{equation} X_n' = X_n - \sum\limits_{i = n+1}^\infty a_i \delta M_i 1_{\|X_{i-1} - x_\star \|\leq r} = X_n - \sum\limits_{i = n+1}^\infty a_i \delta M_i. \end{equation} | (2.35) |
Since there are only finitely many truncations, there exists N\geq n_ \epsilon , such that for all n\geq N , P_{n+1} = 0 , as the truncations have ceased. Consequently, for n\geq N ,
\begin{equation} X'_{n+1} = X'_n - a_{n+1}f(X_n) \end{equation} | (2.36) |
By (2.34) and (2.35), for n\geq N , \|X_n - X_n'\|\leq \epsilon . Since \epsilon > 0 may be arbitrarily small, it will be sufficient to prove X_n'\to x_\star .
4. To obtain convergence of X_n' , we first examine \|X_{n+1}'-x_\star\| . For n\geq N ,
\begin{equation} \begin{split} \|X_{n+1}' - x_\star\|^2 &\leq \|X_{n}' - x_\star\|^2-2a_{n+1}\left\langle {X_n' - x_\star},{f(X_n)}\right\rangle + a_{n+1} \epsilon \bar f^2, \end{split} \end{equation} | (2.37) |
Now consider two cases of this expression. First, assume \|X_n' - x_\star\|\leq\eta . In this case, using (2.33),
\begin{equation} \begin{split} -2a_{n+1}\left\langle {X_n' - x_\star},{f(X_n)}\right\rangle + a_{n+1} \epsilon \bar f^2&\leq a_{n+1}(2\eta \bar f + \epsilon \bar f^2) \\ & \lt a_{n+1}(4R\bar f + \bar f^2) = a_{n+1} B. \end{split} \end{equation} | (2.38) |
where B > 0 is a constant depending only on R and \bar f . For \|X_n' - x_\star\| > \eta , using (2.33)
\begin{equation} \begin{split} \left\langle {X'_{n}-x_\star},{f(X_n) }\right\rangle & = \left\langle {X_{n}-x_\star},{f(X_n) }\right\rangle + \left\langle {X'_{n}-X_n},{f(X_n) }\right\rangle\\ &\geq \left\langle {X_{n}-x_\star},{f(X_n) }\right\rangle - \epsilon \bar{f}\\ & \gt \left\langle {X_{n}-x_\star},{f(X_n) }\right\rangle -\tfrac{1}{2}\delta. \end{split} \end{equation} | (2.39) |
By (2.33),
\begin{equation*} \|X_{n}-x_\star\|\geq \|X'_{n}-x_\star\| -\| X_{n}-X_n'\| \gt \eta - \epsilon \gt \tfrac{1}{2}\eta \end{equation*} |
Since \|X_n - x_\star\| < r too, (2.32) and (2.39) yield the estimate
\begin{equation*} \left\langle {X'_{n}-x_\star},{f(X_n) }\right\rangle \gt \delta - \epsilon \bar{f} \gt \tfrac{1}{2}\delta \end{equation*} |
Thus, in this regime, using (2.33),
\begin{equation} \begin{split} -2a_{n+1}\left\langle {X_n' - x_\star},{f(X_n)}\right\rangle + a_{n+1} \epsilon \bar f^2&\leq -a_{n+1}(\delta - \epsilon \bar f^2)\\ & \lt -\tfrac{1}{2}\delta a_{n+1} = - A a_{n+1} \end{split} \end{equation} | (2.40) |
where A > 0 is a constant depending only on \delta .
Combining estimates (2.38) and (2.40), we can write for n\geq N
\begin{equation} \|X_{n+1}' - x_\star\|^2 \lt \|X_{n}' - x_\star\|^2 - a_{n+1} A 1_{\|X_n' - x_\star\| \gt \eta} + a_{n+1} B 1_{\|X_n' - x_\star\|\leq \eta}. \end{equation} | (2.41) |
5. We now show that \|X_n' - x_\star\|\leq \eta i.o. The argument is by contradiction. Let M\geq N be such that for all n\geq M , \|X_n' - x_\star\| > \eta . For such n ,
\begin{equation} \begin{split} \eta^2 \lt \|X_{n+1}'-x_\star\|^2 & \lt \|X_{n}'-x_\star\|^2 - a_{n+1}A \\ & \lt \|X_{n-1}'-x_\star\|^2 - a_{n+1}A - a_n A\\ & \lt \ldots \lt \|X_M' - x_\star\|^2 - A \sum\limits_{i = M}^n a_{i+1}. \end{split} \end{equation} | (2.42) |
Using Assumption 4 and taking n\to \infty , we obtain a contradiction.
6. Finally, we prove convergence of X_n'\to x_\star . Since X_n'\in B_{\eta}(x_\star) i.o., let
\begin{equation} N' = \inf \{n\geq N\mid \|X_n' - x_\star\| \lt \eta \}. \end{equation} | (2.43) |
For n\geq N' , we can then define
\begin{equation} \varphi(n) = \max\left\{{p\leq n\mid \left \|{X_p' - x_\star}\right\| \lt \eta }\right\}. \end{equation} | (2.44) |
For all such n , \varphi(n)\leq n , and X_{\varphi(n)}'\in B_\eta(x_\star) .
We claim that for n\geq N' ,
\|X_{n+1}' - x_\star\|^2 \lt \|X_{\varphi(n)}' -x_\star\|^2 + B a_{\varphi(n)+1} \lt \eta^2 + Ba_{\varphi(n)+1}. |
First, if n = \varphi(n) , this trivially holds in (2.41). Suppose now that n > \varphi(n) . Then for i = \varphi(n)+1, \varphi(n)+2, \ldots n , \|X_{i}' - x_\star\| > \eta . Consequently,
\begin{equation*} \begin{split} \|X_{n+1}' - x_\star\|^2& \lt \|X_{n}' - x_\star\|^2 \lt \|X_{n-1}' - x_\star\|^2 \lt \ldots\\ & \lt \|X_{\varphi(n)+1}' - x_\star\|^2 \lt \|X_{\varphi(n)}' - x_\star\|^2 + B a_{\varphi(n)+1}\\ & \lt \eta^2 + B a_{\varphi(n)+1} \end{split} \end{equation*} |
As \varphi(n)\to \infty ,
\begin{equation*} \limsup\limits_{n\to \infty }\|X_{n+1}' - x_\star\|^2\leq \eta^2 \end{equation*} |
Since \eta may be arbitrarily small, we conclude that
\limsup\limits_{n\to \infty }\|X_{n+1}' - x_\star\| = \lim\limits_{n\to \infty }\|X_{n+1}' - x_\star\| = 0, |
completing the proof.
Recall from the introduction that our distribution of interest, \mu , is posed on the Borel subsets of Hilbert space \mathcal{H} . We assume that \mu \ll \mu_0 , where \mu_0 = N(m_0, C_0) is some reference Gaussian. Thus, we write
\begin{equation} \frac{d\mu}{d\mu_0} = \frac{1}{Z_\mu}\exp\left\{{-\Phi_\mu(u)}\right\}, \end{equation} | (3.1) |
where \Phi_\nu: X\to \mathbb{R} , X a Banach space, a subspace of \mathcal{H} , of full measure with respect to \mu_0 , a Gaussian on \mathcal{H} , assumed to be continuous. Z_\mu = \mathbb{E}^{\mu_0}[\exp\left\{{-\Phi(u)}\right\}]\in (0, \infty) is the partition function ensuring we have a probability measure.
Let \nu = N(m, C) , be another Gaussian, equivalent to \mu_0 , such that we can write
\begin{equation} \frac{d\nu}{d\mu_0} = \frac{1}{Z_\nu}\exp\left\{{-\Phi_\nu(v)}\right\}, \end{equation} | (3.2) |
Assuming that \nu \ll \mu , we can write
\begin{equation} \mathcal{R}(\nu||\mu) = \mathbb{E}^{\nu}[\Phi_\mu(u) - \Phi_\nu(u)] + \log(Z_\mu) - \log(Z_\nu) \end{equation} | (3.3) |
The assumption that \nu \ll \mu implies that \nu and \mu are equivalent measures. As was proven in [16], if \mathcal{A} is a set of Gaussian measures, closed under weak convergence, such that at least one element of \mathcal{A} is absolutely continuous with respect to \mu , then any minimizing sequence over \mathcal{A} will have a weak subsequential limit.
If we assume, for this work, that C = C_0 , then, by the Cameron-Martin formula (see [9]),
\begin{equation} \Phi_\nu(u) = -\left\langle {u-m},{m -m_0}\right\rangle_{ \mathcal{H}^1} - \frac{1}{2}\left \|{m - m_0}\right\|_{ \mathcal{H}^1}^2, \quad Z_{\nu} = 1. \end{equation} | (3.4) |
Here, \left\langle {\bullet}, {\bullet}\right\rangle_{ \mathcal{H}^1} and \|\bullet\|_{ \mathcal{H}^1} are the inner product and norms of the Cameron-Martin Hilbert space, denoted \mathcal{H}^1 ,
\begin{equation} \left\langle {f},{g}\right\rangle_{ \mathcal{H}^1} = \left\langle {C_0^{-1/2} f},{C_0^{-1/2} g}\right\rangle, \quad \left \|{f}\right\|_{ \mathcal{H}^1}^2 = \left\langle {f},{f}\right\rangle_{ \mathcal{H}^1}^2. \end{equation} | (3.5) |
Convergence to the minimizer will be established in \mathcal{H}^1 , and \mathcal{H}^1 will be the relevant Hilbert space in our application of Theorems 2.1 and 2.2 to this problem.
Letting \nu_0 = N(0, C_0) and v\sim \nu_0 , we can then rewrite (3.3) as
\begin{equation} J(m) \equiv \mathcal{R}(\nu||\mu) = \mathbb{E}^{\nu_0}[\Phi_{\mu}(v + m)] + \frac{1}{2}\left \|{m - m_0}\right\|_{ \mathcal{H}^1}^2 + \log(Z_\mu) \end{equation} | (3.6) |
The Euler-Lagrange equation associated with (3.6), and the second variation, are:
\begin{align} J'(m) & = \mathbb{E}^{\nu_0}[\Phi'_\mu(v+m)] + C_0^{-1}(m-m_0), \end{align} | (3.7) |
\begin{align} J''(m) & = \mathbb{E}^{\nu_0}[\Phi''_\mu(v+m)] + C_0^{-1}. \end{align} | (3.8) |
In [15], it was suggested that rather than try to find a root of (3.7), the equation first be preconditioned by multiplying by C_0 ,
\begin{equation} C_0 \mathbb{E}^{\nu_0}[\Phi'_\mu(v+m)] + (m-m_0), \end{equation} | (3.9) |
and a root of this mapping is sought, instead. Defining
\begin{align} f(m) & = C_0 \mathbb{E}^{\nu_0}[\Phi'_\mu(v+m)] + (m-m_0), \end{align} | (3.10) |
\begin{align} F(m,v) & = C_0\Phi'_\mu(v+m) + (m-m_0). \end{align} | (3.11) |
The Robbins-Monro formulation is then
\begin{equation} m_{n+1} = m_n - a_{n+1} F(m_n, v_{n+1}) + P_{n+1}, \end{equation} | (3.12) |
with v_n \sim \nu_0 , i.i.d.
We thus have
Theorem 3.1. Assume:
● There exists \nu = N(m, C_0)\sim \mu_0 such that \nu\ll\mu .
● \Phi_\mu' and \Phi_\mu'' exist for all u \in \mathcal{H}^1 .
● There exists m_\star , a local minimizer of J , such that J'(m_\star) = 0 .
● The mapping
\begin{equation} m\mapsto \mathbb{E}^{\nu_0}\left[{\left \|{\sqrt{C_0}\Phi_\mu'(m+v)}\right\|^2}\right] \end{equation} | (3.13) |
is bounded on bounded subsets of \mathcal{H}^1 .
● There exists a convex neighborhood U_\star of m_\star and a constant \alpha > 0 , such that for all m\in U_\star , for all u \in \mathcal{H}^1 ,
\begin{equation} \left\langle {J''(m)u},{u}\right\rangle\geq \alpha \left \|{u}\right\|_{ \mathcal{H}^1}^2 \end{equation} | (3.14) |
Then, choosing a_n according to Assumption 4,
● If the subset U_\star can be taken to be all of \mathcal{H}^1 , for the expanding truncation algorithm, m_n \to m_\star a.s. in \mathcal{H}^1 .
● If the subset U_\star is not all of \mathcal{H}^1 , then, taking U_1 to be a bounded (in \mathcal{H}^1 ) convex subset of U_\star , with m_\star \in U_1 , and U_0 any subset of U_1 such that there exist R_0 < R_1 with
U_0 \subset B_{R_0}(x_\star) \subset B_{R_1}(x_\star)\subset U_1, |
for the fixed truncation algorithm, m_n\to m_\star a.s. in \mathcal{H}^1 .
Proof. We split the proof into 2 steps:
1. By the assumptions of the theorem, we clearly satisfy Assumptions 1 and 4. To satisfy Assumption 3, we observe that
\begin{equation*} \mathbb{E}^{\nu_0}[\left \|{F(m,v)}\right\|^2_{ \mathcal{H}^1}]\leq 2 \mathbb{E}^{\nu_0}\left[{\left \|{\sqrt{C_0}\Phi_\mu'(m+v)}\right\|^2}\right] + 2\left \|{m-m_0}\right\|_{ \mathcal{H}^1}^2. \end{equation*} |
This is bounded on bounded subsets of \mathcal{H}^1 .
2. Per the convexity assumption, (3.14), implies Assumption 2, since, by the mean value theorem in function spaces,
\begin{equation*} \begin{split} \left\langle {m-m_\star},{f(m)}\right\rangle_{ \mathcal{H}^1} & = \left\langle {m-m_\star},{C_0\left[{J'(m_\star) +J''(\tilde m)(m-m_\star) }\right]}\right\rangle_{ \mathcal{H}^1}\\ & = \left\langle {m-m_\star},{J''(\tilde m)(m-m_\star)}\right\rangle\geq \alpha\left \|{m-m_\star}\right\|_{ \mathcal{H}^1}^2 \end{split} \end{equation*} |
where \tilde m is some intermediate point between m and m_\star . This completes the proof.
While condition (3.14) is sufficient to obtain convexity, other conditions are possible. For instance, suppose there is a convex open set U_\star containing m_\star and constant \theta\in [0, 1) , such that for all m \in U_\star ,
\begin{equation} \inf\limits_{\substack{u\in \mathcal{H}\\ u\neq 0}} \frac{\left\langle { \mathbb{E}^{\nu_0}[\Phi''_\mu(v+m)]u},{u}\right\rangle}{\left \|{u}\right\|^2}\geq -\theta\lambda_1^{-1}, \end{equation} | (3.15) |
where \lambda_1 is the principal eigenvalue of C_0 . Then this would also imply Assumption 2, since
\begin{equation*} \begin{split} \left\langle {m-m_\star},{f(m)}\right\rangle_{ \mathcal{H}^1} & = \left\langle {m-m_\star},{C_0\left[{J'(m_\star) +J''(\tilde m)(m-m_\star) }\right]}\right\rangle_{ \mathcal{H}^1}\\ & = \left\langle {m-m_\star},{J''(\tilde m)(m-m_\star)}\right\rangle\\ &\geq \left \|{m-m_\star}\right\|_{ \mathcal{H}^1}^2 + \left\langle {m-m_\star},{ \mathbb{E}^{\nu_0}[\Phi''_\mu(v+\tilde m)] (m-m_\star)}\right\rangle\\ &\geq \left \|{m-m_\star}\right\|_{ \mathcal{H}^1}^2 -\theta \lambda_1^{-1} \left \|{m-m_\star}\right\|^2\\ &\geq (1-\theta)\left \|{m-m_\star}\right\|_{ \mathcal{H}^1}^2. \end{split} \end{equation*} |
We mention (3.15) as there may be cases, shown below, for which the operator \mathbb{E}^{\nu_0}[\Phi''_\mu(v+ m)] is obviously nonnegative.
To apply the Robbins-Monro algorithm to the relative entropy minimization problem, the \Phi_\mu functional of interest must be examined. In this section we present a few examples, based on those presented in [15], and examine when the assumptions hold. The one outstanding assumption that we must make is that, a priori, \mu_0 is an equivalent measure to \mu .
Taking \mu_0 = N(0, 1) , the standard unit Gaussian, let V: \mathbb{R} \to \mathbb{R} be a smooth function such that
\begin{equation} \frac{d\mu}{d\mu_0} = \frac{1}{Z_\mu} \exp\left\{{-{ \epsilon^{-1}}V(x)}\right\} \end{equation} | (4.1) |
is a probability measure on \mathbb{R} . For these scalar cases, we use x in place of v . In the above framework,
\begin{align*} F(x,\xi) & = { \epsilon^{-1}}V'(x+\xi) -\xi,\\ f(x) & = { \epsilon^{-1}} \mathbb{E}[V'(x+\xi)]m \\ \Phi_\mu'(x) & = { \epsilon^{-1}}V'(x), \\ \Phi_\mu''(x)& = { \epsilon^{-1}}V''(x) \end{align*} |
and \xi\sim N(0, 1) = \nu_0 = \mu_0 .
Consider the case that
\begin{equation} V(x) = \tfrac{1}{2}x^2 + \tfrac{1}{4}x^4. \end{equation} | (4.2) |
In this case
\begin{align*} F(x,\xi)& = { \epsilon^{-1}}\left({x+\xi + (x+\xi)^3}\right) + x,\\ f(x) & = { \epsilon^{-1}}\left({4x + x^3}\right) +x, \\ \mathbb{E}[\Phi''_\mu(x+\xi)] & = { \epsilon^{-1}}(4 + 3x^2),\\ \mathbb{E}[\left |{\Phi'_\mu(x+\xi)}\right |^2] & = { \epsilon^{-1}} \left({22 + 58 x^2 + 17 x^4 + x^6}\right). \end{align*} |
Since \mathbb{E}[\Phi''_\mu(x+\xi)] \geq 4 { \epsilon^{-1}} , all of our assumptions are satisfied and the expanding truncation algorithm will converge to the unique root at x_\star = 0 a.s. See Figure 1 for an example of the convergence at \epsilon = 0.1 , U_{n} = (-n -1, n+1) , and always restarting at 0.5 .
We refer to this as a "globally convex'' problem since \mathcal{R} is globally convex about the minimizer.
In contrast to the above problem, some mimizers are only "locally'' convex. Consider the case the double well potential
\begin{equation} V(x) = \tfrac{1}{4}(4-x^2)^2 \end{equation} | (4.3) |
Now, the expressions for RM are
\begin{align*} F(x,\xi) & = { \epsilon^{-1}}\left({(x+\xi)^3-4(x+\xi))}\right) + x,\\ f(x) & = { \epsilon^{-1}}\left({x^3-x}\right) +x, \\ \mathbb{E}[\Phi''_\mu(x+\xi)] & = { \epsilon^{-1}}\left({3x^2-1}\right),\\ \mathbb{E}[\left |{\Phi'_\mu(x+\xi)}\right |^2] & = { \epsilon^{-1}} (1 + x^2) (7 + 6 x^2 +x^4). \end{align*} |
In this case, f(x) vanishes at 0 and \pm \sqrt{1- \epsilon} , and J'' changes sign from positive to negative when x enters ({-\sqrt{(1- \epsilon)/3}, \sqrt{({1- \epsilon})/{3}}}) . We must therefore restrict to a fixed trust region if we want to ensure convergence to either of \pm\sqrt{1- \epsilon} .
We ran the problem at \epsilon = 0.1 in two cases. In the first case, U_1 = (0.6, 3.0) and the process always restarts at 2 . This guarantees convergence since the second variation will be strictly postive. In the second case, U_1 = (-0.5, 1.5) , and the process always restarts at -0.1 . Now, the second variation can change sign. The results of these two experiments appear in Figure 2. For some random number sequences the algorithm still converged to \sqrt{1- \epsilon} , even with the poor choice of trust region.
Take \mu_0 = N(m_0(t), C_0) , with
\begin{equation} C_0 = \left({-\frac{d^2}{dt^2}}\right)^{-1}, \end{equation} | (4.4) |
equipped with Dirichlet boundary conditions on \mathcal{H} = L^2(0, 1) .* In this case the Cameron-Martin space \mathcal{H}^1 = H^1_0(0, 1) , the standard Sobolev space equipped with the Dirichlet norm. Let us assume m_0 \in H^1(0, 1) , taking values in \mathbb{R}^d .
* This is the covariance of the standard unit Brownian bridge, Y_t = B_t - t B_1 .
Consider the path space distribution on L^2(0, 1) , induced by
\begin{equation} \frac{d\mu}{d\mu_0} = - \frac{1}{Z_\mu}\exp\left\{{-\Phi_\mu(v)}\right\}, \quad \Phi_\mu(u) = { \epsilon^{-1}}\int_0^1 V(v(t))dt, \end{equation} | (4.5) |
where V: \mathbb{R}^d\to \mathbb{R} is a smooth function. We assume that V is such that this probability distribution exists and that \mu \sim \mu_0 , our reference measure.
We thus seek an \mathbb{R}^d valued function m(t) \in H^1(0, 1) for our Gaussian approximation of \mu , satisfying the boundary conditions
\begin{equation} m(0) = m_-,\quad m(1) = m_+. \end{equation} | (4.6) |
For simplicity, take m_0 = (1-t)m_- + t m_+ , the linear interpolant between m_\pm . As above, we work in the shifted coordinated x(t) = m(t) - m_0(t)\in H^1_0(0, 1) .
Given a path v(t)\in H^1_0 , by the Sobolev embedding, v is continuous with its L^\infty norm controlled by its H^1 norm. Also recall that for \xi \sim N(0, C_0) , in the case of \xi(t) \in \mathbb{R} ,
\begin{equation} \mathbb{E}\left[{\xi(t)^p}\right] = \begin{cases} 0, & \text{$p$ odd},\\ (p-1)!!\left[{t(1-t)}\right]^{\frac p 2}, & \text{$p$ even}. \end{cases} \end{equation} | (4.7) |
Letting \lambda_1 = 1/\pi^2 be the ground state eigenvalue of C_0 ,
\begin{equation*} \begin{split} \mathbb{E}[\|\sqrt{C_0}\Phi'_\mu(v +m_0+\xi)\|^2]&\leq {\lambda_1} \mathbb{E}[\|\Phi'_\mu(v +m_0+\xi)\|^2]\\ &\quad = {\lambda_1}{ \epsilon^{-2}}\int_0^1 \mathbb{E}[{\left |{V'(v(t)+m_0(t)+\xi(t))}\right |^2}]dt. \end{split} \end{equation*} |
The terms involving v+m_0 in the integrand can be controlled by the L^\infty norm, which in turn is controlled by the H^1 norm, while the terms involving \xi can be integrated according to (4.7). As a mapping applied to x , this expression is bounded on bounded subsets of H^1 .
Minimizers will satisfy the ODE
\begin{equation} { \epsilon}^{-1} \mathbb{E}\left[{V'(x+m_0 +\xi)}\right] -x'' = 0,\quad x(0) = x(1) = 0. \end{equation} | (4.8) |
With regard to convexity about a minimizer, m_\star , if, for instance, V'' were pointwise positive definite, then the problem would satisfy (3.15), ensuring convergence. Consider the quartic potential V given by (4.2). In this case,
\begin{equation} \Phi(v) = { \epsilon}^{-1}\int_0^1 \frac{1}{2}v(t)^2 +\frac{1}{4}v(t)^4 dt, \end{equation} | (4.9) |
and
\begin{align*} \Phi'(v+m_0+ \xi) & = { \epsilon}^{-1}\left[{(v+m_0 +\xi) +3(v+m_0 +\xi)^3 }\right]\\ \Phi''(v+m_0 + \xi) & = { \epsilon}^{-1}\left[{1 +3 (v+m_0+\xi)^2}\right],\\ \mathbb{E}[\Phi'(v+m_0 + \xi)]& = { \epsilon}^{-1}\left[{v+m_0 +(v+m_0)^3+ 3 t(1-t) (v+m_0)}\right]\\ \mathbb{E}[\Phi''(v+m_0 + \xi)]& = { \epsilon}^{-1}\left[{1 + 3 (v+m_0)^2 + 3 t(1-t) }\right] \end{align*} |
Since \Phi''(v+m_0+\xi)\geq \epsilon^{-1} , we are guaranteed convergence using expanding trust regions. Taking \epsilon = 0.01 , m_- = 0 and m_+ = 2 , this is illustrated in Figure 3, where we have also solved (4.8) by ODE methods for comparison. As trust regions, we take
\begin{equation} U_n = \left\{{m \in H^1_0(0,1)\mid \left \|{x}\right\|_{H^1}\leq 10+n}\right\}, \end{equation} | (4.10) |
and we always restart at the zero solution Figure 3 also shows robustness to discretization; the number of truncations is relatively insensitive to \Delta t .
For many problems of interest, we do not have global convexity. Consider the double well potential (4.3), but in the case of paths,
\begin{equation} \Phi(u) = { \epsilon^{-1}}\int_0^1\frac{1}{4} (4-v(t)^2)^2dt. \end{equation} | (4.11) |
Then,
\begin{align*} \Phi'(v + m_0 + \xi)& = { \epsilon}^{-1}\left[{(v + m_0 + \xi)^3 - 4 (v + m_0 + \xi)}\right]\\ \Phi''(v + m_0 + \xi) & = { \epsilon}^{-1}\left[{3 (v + m_0 + \xi)^2 - 4}\right],\\ \mathbb{E}[\Phi'(v+m_0 + \xi)]& = { \epsilon}^{-1}\left[{(v+m_0)^3 + 3 t(1-t) (v+m_0)-4(v+m_0)}\right]\\ \mathbb{E}[\Phi''(v+m_0 + \xi)]& = { \epsilon}^{-1}\left[{3(v+m_0)^2 + 3 t(1-t) -4}\right] \end{align*} |
Here, we take m_- = 0 , m_+ = 2 , and \epsilon = 0.01 . We have plotted the numerically solved ODE in Figure 4. Also plotted is \mathbb{E}[\Phi''(v_\star +m_0+ \xi)] . Note that \mathbb{E}[\Phi''(v_\star +m_0+ \xi)] is not sign definite, becoming as small as -400 . Since C_0 has \lambda_1 = 1/\pi^2 \approx 0.101 , (3.15) cannot apply.
Discretizing the Schrödinger operator
\begin{equation} J''(v_\star) = -\frac{d^2}{dt^2} + { \epsilon}^{-1}\left({3(v_\star(t)+m_0(t))^2 + 3 t(1-t) -4}\right), \end{equation} | (4.12) |
we numerically compute the eigenvalues. Plotted in Figure 5, we see that the minimal eigenvalue of J''(m_\star) is approximately \mu_1\approx 550 . Therefore,
\begin{equation} \left\langle {J''(x_\star)u},{u}\right\rangle\geq \mu_1 \left \|{u}\right\|^2_{L^2}\Rightarrow \left\langle {J''(x)u},{u}\right\rangle\geq \alpha\left \|{u}\right\|_{H^1}^2, \end{equation} | (4.13) |
for all v in some neighborhood of v_\star . For an appropriately selected fixed trust region, the algorithm will converge.
However, we can show that the convexity condition is not global. Consider the path m(t) = 2t^2 , which satisfies the boundary conditions. As shown in Figure 5, this path induces negative eigenvalues.
Despite this, we are still observe convergence. Using the fixed trust region
\begin{equation} U_1 = \left\{{x\in H^1_0(0,1)\mid \left \|{x}\right\|_{H^1}\leq 100}\right\}, \end{equation} | (4.14) |
we obtain the results in Figure 6. Again, the convergence is robust to discretization.
We have shown that the Robbins-Monro algorithm, with both fixed and expanding trust regions, can be applied to Hilbert space valued problems, adapting the finite dimensional proof of [12]. We have also constructed sufficient conditions for which the relative entropy minimization problem fits within this framework.
One problem we did not address here was how to identify fixed trust regions. Indeed, that requires a tremendous amount of a priori information that is almost certainly not available. We interpret that result as a local convergence result that gives a theoretical basis for applying the algorithm. In practice, since the root is likely unknown, one might run some numerical experiments to identify a reasonable trust region, or just use expanding trust regions. The practitioner will find that the algorithm converges to a solution, though perhaps not the one originally envisioned. A more sophisticated analysis may address the convergence to a set of roots, while being agnostic as to which zero is found.
Another problem we did not address was how to optimize not just the mean, but also the covariance in the Gaussian. As discussed in [15], it is necessary to parameterize the covariance in some way, which will be application specific. Thus, while the form of the first variation of relative entropy with respect to the mean, (3.7), is quite generic, the corresponding expression for the covariance will be specific to the covariance parameterization. Additional constraints are also necessary to guarantee that the parameters always induce a covariance operator. We leave such specialization as future work.
This work was supported by US Department of Energy Award DE-SC0012733. This work was completed under US National Science Foundation Grant DMS-1818716. The authors would like to thank J. Lelong for helpful comments, along with anonymous reviewers whose reports significantly impacted our work.
The authors declare that there is no conflicts of interest in this paper.
[1] |
Cheung AK, Sarnak MJ, Yan G, et al. (2004) Cardiac diseases in maintenance hemodialysis patients: results of the HEMO Study. Kidney Int 65: 2380-2389. doi: 10.1111/j.1523-1755.2004.00657.x
![]() |
[2] |
Foley RN, Parfrey PS, Harnett JD, et al. (1995) Clinical and echocardiographic disease in patients starting end-stage renal disease therapy. Kidney Int 47: 186-192. doi: 10.1038/ki.1995.22
![]() |
[3] | ANZDATA Registry. 38th Report, Chapter 3: Mortality in End Stage Kidney Disease. Australia and New Zealand Dialysis and Transplant Registry, Adelaide, Australia. 2016. |
[4] |
Foley RN, Parfrey PS, Sarnak MJ (1998) Clinical epidemiology of cardiovascular disease in chronic renal disease. Am J Kidney Dis 32: S112-119. doi: 10.1053/ajkd.1998.v32.pm9820470
![]() |
[5] |
Longenecker JC, Coresh J, Powe NR, et al. (2002) Traditional cardiovascular disease risk factors in dialysis patients compared with the general population: the CHOICE Study. J Am Soc Nephrol 13: 1918-1927. doi: 10.1097/01.ASN.0000019641.41496.1E
![]() |
[6] |
Ruilope LM, van Veldhuisen DJ, Ritz E, et al. (2001) Renal function: the Cinderella of cardiovascular risk profile. J Am Coll Cardiol 38: 1782-1787. doi: 10.1016/S0735-1097(01)01627-8
![]() |
[7] |
Agarwal R (2005) Hypertension in chronic kidney disease and dialysis: pathophysiology and management. Cardiol Clin 23: 237-248. doi: 10.1016/j.ccl.2005.03.003
![]() |
[8] |
Neumann J, Ligtenberg G, Klein II, et al. (2004) Sympathetic hyperactivity in chronic kidney disease: pathogenesis, clinical relevance, and treatment. Kidney Int 65: 1568-1576. doi: 10.1111/j.1523-1755.2004.00552.x
![]() |
[9] |
Covic A, Kothawala P, Bernal M, et al. (2009) Systematic review of the evidence underlying the association between mineral metabolism disturbances and risk of all-cause mortality, cardiovascular mortality and cardiovascular events in chronic kidney disease. Nephrol Dial Transplant 24: 1506-1523. doi: 10.1093/ndt/gfn613
![]() |
[10] |
Kaysen GA, Eiserich JP (2004) The role of oxidative stress-altered lipoprotein structure and function and microinflammation on cardiovascular risk in patients with minor renal dysfunction. J Am Soc Nephrol 15: 538-548. doi: 10.1097/01.ASN.0000111744.00916.E6
![]() |
[11] |
Wang AY, Lam CW, Chan IH, et al. (2010) Sudden cardiac death in end-stage renal disease patients: a 5-year prospective analysis. Hypertension 56: 210-216. doi: 10.1161/HYPERTENSIONAHA.110.151167
![]() |
[12] | Martinez-Rumayor A, Richards AM, Burnett JC, et al. (2008) Biology of the natriuretic peptides. Am J Cardiol 101: 3-8. |
[13] | Mukoyama M, Nakao K, Hosoda K, et al. (1991) Brain natriuretic peptide as a novel cardiac hormone in humans. Evidence for an exquisite dual natriuretic peptide system, atrial natriuretic peptide and brain natriuretic peptide. J Clin Invest 87: 1402-1412. |
[14] |
Yasue H, Yoshimura M, Sumida H, et al. (1994) Localization and mechanism of secretion of B-type natriuretic peptide in comparison with those of A-type natriuretic peptide in normal subjects and patients with heart failure. Circulation 90: 195-203. doi: 10.1161/01.CIR.90.1.195
![]() |
[15] | Gerbes AL, Dagnino L, Nguyen T, et al. (1994) Transcription of brain natriuretic peptide and atrial natriuretic peptide genes in human tissues. J Clin Endocrinol Metab 78: 1307-1311. |
[16] | Kinnunen P, Vuolteenaho O, Ruskoaho H (1993) Mechanisms of atrial and brain natriuretic peptide release from rat ventricular myocardium: effect of stretching. Endocrinology 132: 1961-1970. |
[17] | Bruneau BG, Piazza LA, de Bold AJ (1997) BNP gene expression is specifically modulated by stretch and ET-1 in a new model of isolated rat atria. Am J Physiol 273: H2678-2686. |
[18] |
Liang F, Gardner DG (1999) Mechanical strain activates BNP gene transcription through a p38/NF-kappaB-dependent mechanism. J Clin Invest 104: 1603-1612. doi: 10.1172/JCI7362
![]() |
[19] |
Bibbins-Domingo K, Ansari M, Schiller NB, et al. (2003) B-type natriuretic peptide and ischemia in patients with stable coronary disease: data from the Heart and Soul study. Circulation 108: 2987-2992. doi: 10.1161/01.CIR.0000103681.04726.9C
![]() |
[20] |
de Bold AJ (2009) Cardiac natriuretic peptides gene expression and secretion in inflammation. J Investig Med 57: 29-32. doi: 10.2310/JIM.0b013e3181948b37
![]() |
[21] | Bruneau BG, Piazza LA, de Bold AJ (1996) Alpha 1-adrenergic stimulation of isolated rat atria results in discoordinate increases in natriuretic peptide secretion and gene expression and enhances Egr-1 and c-Myc expression. Endocrinology 137: 137-143. |
[22] |
Wiese S, Breyer T, Dragu A, et al. (2000) Gene expression of brain natriuretic peptide in isolated atrial and ventricular human myocardium: influence of angiotensin II and diastolic fiber length. Circulation 102: 3074-3079. doi: 10.1161/01.CIR.102.25.3074
![]() |
[23] |
Maisel AS, Krishnaswamy P, Nowak RM, et al. (2002) Rapid measurement of B-type natriuretic peptide in the emergency diagnosis of heart failure. N Engl J Med 347: 161-167. doi: 10.1056/NEJMoa020233
![]() |
[24] |
McCullough PA, Nowak RM, McCord J, et al. (2002) B-type natriuretic peptide and clinical judgment in emergency diagnosis of heart failure: analysis from Breathing Not Properly (BNP) Multinational Study. Circulation 106: 416-422. doi: 10.1161/01.CIR.0000025242.79963.4C
![]() |
[25] |
Doust JA, Pietrzak E, Dobson A, et al. (2005) How well does B-type natriuretic peptide predict death and cardiac events in patients with heart failure: systematic review. Bmj 330: 625. doi: 10.1136/bmj.330.7492.625
![]() |
[26] |
Troughton RW, Frampton CM, Yandle TG, et al. (2000) Treatment of heart failure guided by plasma aminoterminal brain natriuretic peptide (N-BNP) concentrations. Lancet 355: 1126-1130. doi: 10.1016/S0140-6736(00)02060-2
![]() |
[27] |
Pfisterer M, Buser P, Rickli H, et al. (2009) BNP-guided vs symptom-guided heart failure therapy: the Trial of Intensified vs Standard Medical Therapy in Elderly Patients With Congestive Heart Failure (TIME-CHF) randomized trial. Jama 301: 383-392. doi: 10.1001/jama.2009.2
![]() |
[28] |
Berger R, Moertl D, Peter S, et al. (2010) N-terminal pro-B-type natriuretic peptide-guided, intensive patient management in addition to multidisciplinary care in chronic heart failure a 3-arm, prospective, randomized pilot study. J Am Coll Cardiol 55: 645-653. doi: 10.1016/j.jacc.2009.08.078
![]() |
[29] |
Eurlings LW, van Pol PE, Kok WE, et al. (2010) Management of chronic heart failure guided by individual N-terminal pro-B-type natriuretic peptide targets: results of the PRIMA (Can PRo-brain-natriuretic peptide guided therapy of chronic heart failure IMprove heart fAilure morbidity and mortality?) study. J Am Coll Cardiol 56: 2090-2100. doi: 10.1016/j.jacc.2010.07.030
![]() |
[30] |
Persson H, Erntell H, Eriksson B, et al. (2010) Improved pharmacological therapy of chronic heart failure in primary care: a randomized Study of NT-proBNP Guided Management of Heart Failure--SIGNAL-HF (Swedish Intervention study—Guidelines and NT-proBNP AnaLysis in Heart Failure). Eur J Heart Fail 12: 1300-1308. doi: 10.1093/eurjhf/hfq169
![]() |
[31] |
Lainchbury JG, Troughton RW, Strangman KM, et al. (2009) N-terminal pro-B-type natriuretic peptide-guided treatment for chronic heart failure: results from the BATTLESCARRED (NT-proBNP-Assisted Treatment To Lessen Serial Cardiac Readmissions and Death) trial. J Am Coll Cardiol 55: 53-60. doi: 10.1016/j.jacc.2009.02.095
![]() |
[32] |
Shah MR, Califf RM, Nohria A, et al. (2011) The STARBRITE trial: a randomized, pilot study of B-type natriuretic peptide-guided therapy in patients with advanced heart failure. J Card Fail 17: 613-621. doi: 10.1016/j.cardfail.2011.04.012
![]() |
[33] |
Karlstrom P, Alehagen U, Boman K, et al. (2011) Brain natriuretic peptide-guided treatment does not improve morbidity and mortality in extensively treated patients with chronic heart failure: responders to treatment have a significantly better outcome. Eur J Heart Fail 13: 1096-1103. doi: 10.1093/eurjhf/hfr078
![]() |
[34] |
Januzzi JL, Jr., Rehman SU, Mohammed AA, et al. (2011) Use of amino-terminal pro-B-type natriuretic peptide to guide outpatient therapy of patients with chronic left ventricular systolic dysfunction. J Am Coll Cardiol 58: 1881-1889. doi: 10.1016/j.jacc.2011.03.072
![]() |
[35] |
Jourdain P, Jondeau G, Funck F, et al. (2007) Plasma brain natriuretic peptide-guided therapy to improve outcome in heart failure: the STARS-BNP Multicenter Study. J Am Coll Cardiol 49: 1733-1739. doi: 10.1016/j.jacc.2006.10.081
![]() |
[36] |
Anguita M, Esteban F, Castillo JC, et al. (2010) [Usefulness of brain natriuretic peptide levels, as compared with usual clinical control, for the treatment monitoring of patients with heart failure]. Med Clin (Barc) 135: 435-440. doi: 10.1016/j.medcli.2009.11.048
![]() |
[37] |
Troughton RW, Frampton CM, Brunner-La Rocca HP, et al. (2014) Effect of B-type natriuretic peptide-guided treatment of chronic heart failure on total mortality and hospitalization: an individual patient meta-analysis. Eur Heart J 35: 1559-1567. doi: 10.1093/eurheartj/ehu090
![]() |
[38] |
Mant J, Al-Mohammad A, Swain S, et al. (2011) Management of chronic heart failure in adults: synopsis of the National Institute For Health and clinical excellence guideline. Ann Intern Med 155: 252-259. doi: 10.7326/0003-4819-155-4-201108160-00009
![]() |
[39] |
van Kimmenade RR, Januzzi JL, Jr., Bakker JA, et al. (2009) Renal clearance of B-type natriuretic peptide and amino terminal pro-B-type natriuretic peptide a mechanistic study in hypertensive subjects. J Am Coll Cardiol 53: 884-890. doi: 10.1016/j.jacc.2008.11.032
![]() |
[40] |
Tsai SH, Lin YY, Chu SJ, et al. (2010) Interpretation and use of natriuretic peptides in non-congestive heart failure settings. Yonsei Med J 51: 151-163. doi: 10.3349/ymj.2010.51.2.151
![]() |
[41] | Januzzi JL, van Kimmenade R, Lainchbury J, et al. (2006) NT-proBNP testing for diagnosis and short-term prognosis in acute destabilized heart failure: an international pooled analysis of 1256 patients: the International Collaborative of NT-proBNP Study. Eur Heart J 27: 330-337. |
[42] |
Gutierrez OM, Tamez H, Bhan I, et al. (2008) N-terminal pro-B-type natriuretic peptide (NT-proBNP) concentrations in hemodialysis patients: prognostic value of baseline and follow-up measurements. Clin Chem 54: 1339-1348. doi: 10.1373/clinchem.2007.101691
![]() |
[43] |
Wang AY, Lam CW, Yu CM, et al. (2007) N-terminal pro-brain natriuretic peptide: an independent risk predictor of cardiovascular congestion, mortality, and adverse cardiovascular outcomes in chronic peritoneal dialysis patients. J Am Soc Nephrol 18: 321-330. doi: 10.1681/ASN.2005121299
![]() |
[44] |
Satyan S, Light RP, Agarwal R (2007) Relationships of N-terminal pro-B-natriuretic peptide and cardiac troponin T to left ventricular mass and function and mortality in asymptomatic hemodialysis patients. Am J Kidney Dis 50: 1009-1019. doi: 10.1053/j.ajkd.2007.08.017
![]() |
[45] |
Paniagua R, Amato D, Mujais S, et al. (2008) Predictive value of brain natriuretic peptides in patients on peritoneal dialysis: results from the ADEMEX trial. Clin J Am Soc Nephrol 3: 407-415. doi: 10.2215/CJN.03820907
![]() |
[46] | David S, Kumpers P, Seidler V, et al. (2008) Diagnostic value of N-terminal pro-B-type natriuretic peptide (NT-proBNP) for left ventricular dysfunction in patients with chronic kidney disease stage 5 on haemodialysis. Nephrol Dial Transplant 23: 1370-1377. |
[47] | Parfrey PS (2000) Cardiac disease in dialysis patients: diagnosis, burden of disease, prognosis, risk factors and management. Nephrol Dial Transplant 15 Suppl 5: 58-68. |
[48] |
Fahim MA, Hayen A, Horvath AR, et al. (2015) N-terminal pro-B-type natriuretic peptide variability in stable dialysis patients. Clin J Am Soc Nephrol 10: 620-629. doi: 10.2215/CJN.09060914
![]() |
[49] | Lippi G, Tessitore N, Luca Salvagno G, et al. (2007) Influence of haemodialysis on the NT-proBNP plasma concentration. Clin Chem Lab Med 45: 1414-1415. |
[50] |
Wahl HG, Graf S, Renz H, et al. (2004) Elimination of the cardiac natriuretic peptides B-type natriuretic peptide (BNP) and N-terminal proBNP by hemodialysis. Clin Chem 50: 1071-1074. doi: 10.1373/clinchem.2003.030692
![]() |
[51] |
Clerico A, Caprioli R, Del Ry S, et al. (2001) Clinical relevance of cardiac natriuretic peptides measured by means of competitive and non-competitive immunoassay methods in patients with renal failure on chronic hemodialysis. J Endocrinol Invest 24: 24-30. doi: 10.1007/BF03343804
![]() |
[52] |
Racek J, Kralova H, Trefil L, et al. (2006) Brain natriuretic peptide and N-terminal proBNP in chronic haemodialysis patients. Nephron Clin Pract 103: c162-172. doi: 10.1159/000092914
![]() |
[53] |
Dautin G, Boudjeltia S, Soltani Z, et al. (2007) The changes in NT-proBNP plasma concentrations during dialysis are highly dependent of the dialysis membrane ultrafiltration coefficient. Clin Chim Acta 376: 237-239. doi: 10.1016/j.cca.2006.08.004
![]() |
[54] | Bargnoux AS, Klouche K, Fareh J, et al. (2008) Prohormone brain natriuretic peptide (proBNP), BNP and N-terminal-proBNP circulating levels in chronic hemodialysis patients. Correlation with ventricular function, fluid removal and effect of hemodiafiltration. Clin Chem Lab Med 46: 1019-1024. |
[55] | Sheen V, Bhalla V, Tulua-Tata A, et al. (2007) The use of B-type natriuretic peptide to assess volume status in patients with end-stage renal disease. Am Heart J 153: 244.e241-245. |
[56] |
Flemmer M, Rajab H, Mathena T, et al. (2008) Blood B-type natriuretic peptide and dialysis: present assessment and future analyses. South Med J 101: 1094-1100. doi: 10.1097/SMJ.0b013e318189aa56
![]() |
[57] | Chiarelli G, Beaulieu M, Taylor P, et al. (2011) Elimination of BNP by peritoneal dialysis: investigation of analytical issues. Perit Dial Int 31: 199-202. |
[58] |
Obineche EN, Pathan JY, Fisher S, et al. (2006) Natriuretic peptide and adrenomedullin levels in chronic renal failure and effects of peritoneal dialysis. Kidney Int 69: 152-156. doi: 10.1038/sj.ki.5000009
![]() |
[59] | Granja CA, Tailor PT, Gorban-Brennan N, et al. (2007) Brain natriuretic peptide and impedance cardiography to assess volume status in peritoneal dialysis patients. Adv Perit Dial 23: 155-160. |
[60] | Lee JA, Kim DH, Yoo SJ, et al. (2006) Association between serum n-terminal pro-brain natriuretic peptide concentration and left ventricular dysfunction and extracellular water in continuous ambulatory peritoneal dialysis patients. Perit Dial Int 26: 360-365. |
[61] |
Booth J, Pinney J, Davenport A (2010) N-terminal proBNP--marker of cardiac dysfunction, fluid overload, or malnutrition in hemodialysis patients? Clin J Am Soc Nephrol 5: 1036-1040. doi: 10.2215/CJN.09001209
![]() |
[62] | Gangji AS, Helal BA, Churchill DN, et al. (2008) Association between N-terminal propeptide B-type natriuretic peptide and markers of hypervolemia. Perit Dial Int 28: 308-311. |
[63] |
Jacobs LH, van de Kerkhof JJ, Mingels AM, et al. (2010) Inflammation, overhydration and cardiac biomarkers in haemodialysis patients: a longitudinal study. Nephrol Dial Transplant 25: 243-248. doi: 10.1093/ndt/gfp417
![]() |
[64] |
Lee SW, Song JH, Kim GA, et al. (2003) Plasma brain natriuretic peptide concentration on assessment of hydration status in hemodialysis patient. Am J Kidney Dis 41: 1257-1266. doi: 10.1016/S0272-6386(03)00358-5
![]() |
[65] | Fagugli RM, Palumbo B, Ricciardi D, et al. (2003) Association between brain natriuretic peptide and extracellular water in hemodialysis patients. Nephron Clin Pract 95: c60-66. |
[66] |
Sommerer C, Beimler J, Schwenger V, et al. (2007) Cardiac biomarkers and survival in haemodialysis patients. Eur J Clin Invest 37: 350-356. doi: 10.1111/j.1365-2362.2007.01785.x
![]() |
[67] | Bavbek N, Akay H, Altay M, et al. (2007) Serum BNP concentration and left ventricular mass in CAPD and automated peritoneal dialysis patients. Perit Dial Int 27: 663-668. |
[68] | Zoccali C, Mallamaci F, Benedetto FA, et al. (2001) Cardiac natriuretic peptides are related to left ventricular mass and function and predict mortality in dialysis patients. J Am Soc Nephrol 12: 1508-1515. |
[69] |
Foley RN, Curtis BM, Randell EW, et al. (2010) Left ventricular hypertrophy in new hemodialysis patients without symptomatic cardiac disease. Clin J Am Soc Nephrol 5: 805-813. doi: 10.2215/CJN.07761109
![]() |
[70] |
Mallamaci F, Zoccali C, Tripepi G, et al. (2001) Diagnostic potential of cardiac natriuretic peptides in dialysis patients. Kidney Int 59: 1559-1566. doi: 10.1046/j.1523-1755.2001.0590041559.x
![]() |
[71] |
Choi SY, Lee JE, Jang EH, et al. (2008) Association between changes in N-terminal pro-brain natriuretic peptide levels and changes in left ventricular mass index in stable hemodialysis patients. Nephron Clin Pract 110: c93-100. doi: 10.1159/000157622
![]() |
[72] |
Takase H, Dohi Y, Toriyama T, et al. (2011) B-type natriuretic peptide levels and cardiovascular risk in patients with diastolic dysfunction on chronic haemodialysis: cross-sectional and observational studies. Nephrol Dial Transplant 26: 683-690. doi: 10.1093/ndt/gfq408
![]() |
[73] |
Goto T, Takase H, Toriyama T, et al. (2002) Increased circulating levels of natriuretic peptides predict future cardiac event in patients with chronic hemodialysis. Nephron 92: 610-615. doi: 10.1159/000064100
![]() |
[74] |
Tripepi G, Mattace-Raso F, Mallamaci F, et al. (2009) Biomarkers of left atrial volume: a longitudinal study in patients with end stage renal disease. Hypertension 54: 818-824. doi: 10.1161/HYPERTENSIONAHA.109.136804
![]() |
[75] |
Wang AY, Wang M, Lam CW, et al. (2011) Heart failure in long-term peritoneal dialysis patients: a 4-year prospective analysis. Clin J Am Soc Nephrol 6: 805-812. doi: 10.2215/CJN.07130810
![]() |
[76] |
Winkler K, Wanner C, Drechsler C, et al. (2008) Change in N-terminal-pro-B-type-natriuretic-peptide and the risk of sudden death, stroke, myocardial infarction, and all-cause mortality in diabetic dialysis patients. Eur Heart J 29: 2092-2099. doi: 10.1093/eurheartj/ehn278
![]() |
[77] |
Breidthardt T, Kalbermatter S, Socrates T, et al. (2011) Increasing B-type natriuretic peptide levels predict mortality in unselected haemodialysis patients. Eur J Heart Fail 13: 860-867. doi: 10.1093/eurjhf/hfr057
![]() |
[78] | Doust J (2010) Qualification versus validation of biomarkers. Scand J Clin Lab Invest Suppl 242: 40-43. |
[79] |
Katus HA, Remppis A, Scheffold T, et al. (1991) Intracellular compartmentation of cardiac troponin T and its release kinetics in patients with reperfused and nonreperfused myocardial infarction. Am J Cardiol 67: 1360-1367. doi: 10.1016/0002-9149(91)90466-X
![]() |
[80] |
Turer AT, Addo TA, Martin JL, et al. (2011) Myocardial ischemia induced by rapid atrial pacing causes troponin T release detectable by a highly sensitive assay: insights from a coronary sinus sampling study. J Am Coll Cardiol 57: 2398-2405. doi: 10.1016/j.jacc.2010.11.066
![]() |
[81] |
Hessel MH, Atsma DE, van der Valk EJ, et al. (2008) Release of cardiac troponin I from viable cardiomyocytes is mediated by integrin stimulation. Pflugers Arch 455: 979-986. doi: 10.1007/s00424-007-0354-8
![]() |
[82] |
Thygesen K, Alpert JS, Jaffe AS, et al. (2012) Third universal definition of myocardial infarction. Circulation 126: 2020-2035. doi: 10.1161/CIR.0b013e31826e1058
![]() |
[83] |
Keddis MT, El-Zoghby ZM, El Ters M, et al. (2013) Cardiac troponin T before and after kidney transplantation: determinants and implications for posttransplant survival. Am J Transplant 13: 406-414. doi: 10.1111/j.1600-6143.2012.04317.x
![]() |
[84] | Wu AH, Feng YJ, Roper L, et al. (1997) Cardiac troponins T and I before and after renal transplantation. Clin Chem 43: 411-412. |
[85] |
Fredericks S, Chang R, Gregson H, et al. (2001) Circulating cardiac troponin-T in patients before and after renal transplantation. Clin Chim Acta 310: 199-203. doi: 10.1016/S0009-8981(01)00547-2
![]() |
[86] |
Bozbas H, Korkmaz ME, Atar I, et al. (2004) Serum levels of cardiac enzymes before and after renal transplantation. Clin Cardiol 27: 559-562. doi: 10.1002/clc.4960271007
![]() |
[87] |
Ellis K, Dreisbach AW, Lertora JL (2001) Plasma elimination of cardiac troponin I in end-stage renal disease. South Med J 94: 993-996. doi: 10.1097/00007611-200194100-00011
![]() |
[88] |
Fahie-Wilson MN, Carmichael DJ, Delaney MP, et al. (2006) Cardiac troponin T circulates in the free, intact form in patients with kidney failure. Clin Chem 52: 414-420. doi: 10.1373/clinchem.2005.062307
![]() |
[89] |
Jacobs LH, van de Kerkhof J, Mingels AM, et al. (2009) Haemodialysis patients longitudinally assessed by highly sensitive cardiac troponin T and commercial cardiac troponin T and cardiac troponin I assays. Ann Clin Biochem 46: 283-290. doi: 10.1258/acb.2009.008197
![]() |
[90] |
Kumar N, Michelis MF, DeVita MV, et al. (2011) Troponin I levels in asymptomatic patients on haemodialysis using a high-sensitivity assay. Nephrol Dial Transplant 26: 665-670. doi: 10.1093/ndt/gfq442
![]() |
[91] |
Wolley M, Stewart R, Curry E, et al. (2013) Variation in and prognostic importance of troponin T measured using a high-sensitivity assay in clinically stable haemodialysis patients. Clin Kidney J 6: 402-409. doi: 10.1093/ckj/sfs122
![]() |
[92] |
Hill SA, Cleve R, Carlisle E, et al. (2009) Intra-individual variability in troponin T concentration in dialysis patients. Clin Biochem 42: 991-995. doi: 10.1016/j.clinbiochem.2009.03.027
![]() |
[93] |
Pianta TJ, Horvath AR, Ellis VM, et al. (2012) Cardiac high-sensitivity troponin T measurement: a layer of complexity in managing haemodialysis patients. Nephrology (Carlton) 17: 636-641. doi: 10.1111/j.1440-1797.2012.01625.x
![]() |
[94] | Katus HA, Haller C, Muller-Bardorff M, et al. (1995) Cardiac troponin T in end-stage renal disease patients undergoing chronic maintenance hemodialysis. Clin Chem 41: 1201-1203. |
[95] | Collinson PO, Stubbs PJ, Rosalki SB (1995) Cardiac troponin T in renal disease. Clin Chem 41: 1671-1673. |
[96] | Li D, Jialal I, Keffer J (1996) Greater frequency of increased cardiac troponin T than increased cardiac troponin I in patients with chronic renal failure. Clin Chem 42: 114-115. |
[97] |
Hickman PE, McGill D, Potter JM, et al. (2015) Multiple biomarkers including cardiac troponins T and I measured by high-sensitivity assays, as predictors of long-term mortality in patients with chronic renal failure who underwent dialysis. Am J Cardiol 115: 1601-1606. doi: 10.1016/j.amjcard.2015.02.066
![]() |
[98] | Lippi G, Tessitore N, Montagnana M, et al. (2008) Influence of sampling time and ultrafiltration coefficient of the dialysis membrane on cardiac troponin I and T. Arch Pathol Lab Med 132: 72-76. |
[99] |
Gaze DC, Collinson PO (2014) Cardiac troponin I but not cardiac troponin T adheres to polysulfone dialyser membranes in an in vitro haemodialysis model: explanation for lower serum cTnI concentrations following dialysis. Open Heart 1: e000108. doi: 10.1136/openhrt-2014-000108
![]() |
[100] | Nunes JP, Sampaio S, Cerqueira A, et al. (2015) Anti-troponin I antibodies in renal transplant patients. Rev Port Cardiol 34: 85-89. |
[101] |
Sacchetti A, Harris R, Patel K, et al. (1991) Emergency department presentation of renal dialysis patients: indications for EMS transport directly to dialysis centers. J Emerg Med 9: 141-144. doi: 10.1016/0736-4679(91)90320-F
![]() |
[102] |
McDonald SP, Tong B (2011) Morbidity burden of end-stage kidney disease in Australia: hospital separation rates among people receiving kidney replacement therapy. Nephrology (Carlton) 16: 758-766. doi: 10.1111/j.1440-1797.2011.01486.x
![]() |
[103] |
Herzog CA, Littrell K, Arko C, et al. (2007) Clinical characteristics of dialysis patients with acute myocardial infarction in the United States: a collaborative project of the United States Renal Data System and the National Registry of Myocardial Infarction. Circulation 116: 1465-1472. doi: 10.1161/CIRCULATIONAHA.107.696765
![]() |
[104] |
Sosnov J, Lessard D, Goldberg RJ, et al. (2006) Differential symptoms of acute myocardial infarction in patients with kidney disease: a community-wide perspective. Am J Kidney Dis 47: 378-384. doi: 10.1053/j.ajkd.2005.11.017
![]() |
[105] |
Jaffe AS, Apple FS (2012) The third Universal Definition of Myocardial Infarction--moving forward. Clin Chem 58: 1727-1728. doi: 10.1373/clinchem.2012.196832
![]() |
[106] |
Wu AH, Jaffe AS, Apple FS, et al. (2007) National Academy of Clinical Biochemistry laboratory medicine practice guidelines: use of cardiac troponin and B-type natriuretic peptide or N-terminal proB-type natriuretic peptide for etiologies other than acute coronary syndromes and heart failure. Clin Chem 53: 2086-2096. doi: 10.1373/clinchem.2007.095679
![]() |
[107] | Fahim MA, Hayen AD, Horvath AR, et al. (2015) Biological variation of high sensitivity cardiac troponin-T in stable dialysis patients: implications for clinical practice. Clin Chem Lab Med 53: 715-722. |
[108] |
Michos ED, Wilson LM, Yeh HC, et al. (2014) Prognostic value of cardiac troponin in patients with chronic kidney disease without suspected acute coronary syndrome: a systematic review and meta-analysis. Ann Intern Med 161: 491-501. doi: 10.7326/M14-0743
![]() |
[109] | (2005) K/DOQI clinical practice guidelines for cardiovascular disease in dialysis patients. Am J Kidney Dis 45: S1-153. |
[110] |
Yancy CW, Jessup M, Bozkurt B, et al. (2013) 2013 ACCF/AHA guideline for the management of heart failure: a report of the American College of Cardiology Foundation/American Heart Association Task Force on Practice Guidelines. J Am Coll Cardiol 62: e147-239. doi: 10.1016/j.jacc.2013.05.019
![]() |
[111] |
Mueller T, Leitner I, Egger M, et al. (2015) Association of the biomarkers soluble ST2, galectin-3 and growth-differentiation factor-15 with heart failure and other non-cardiac diseases. Clin Chim Acta 445: 155-160. doi: 10.1016/j.cca.2015.03.033
![]() |
[112] |
Mueller T, Dieplinger B (2013) The Presage((R)) ST2 Assay: analytical considerations and clinical applications for a high-sensitivity assay for measurement of soluble ST2. Expert Rev Mol Diagn 13: 13-30. doi: 10.1586/erm.12.128
![]() |
[113] | Lippi G, Cervellin G (2014) Risk assessment of post-infarction heart failure. Systematic review on the role of emerging biomarkers. Crit Rev Clin Lab Sci 51: 13-29. |
[114] |
Weinberg EO, Shimpo M, De Keulenaer GW, et al. (2002) Expression and regulation of ST2, an interleukin-1 receptor family member, in cardiomyocytes and myocardial infarction. Circulation 106: 2961-2966. doi: 10.1161/01.CIR.0000038705.69871.D9
![]() |
[115] |
Demyanets S, Speidl WS, Tentzeris I, et al. (2014) Soluble ST2 and interleukin-33 levels in coronary artery disease: relation to disease activity and adverse outcome. PLoS One 9: e95055. doi: 10.1371/journal.pone.0095055
![]() |
[116] |
Dieplinger B, Egger M, Haltmayer M, et al. (2014) Increased soluble ST2 predicts long-term mortality in patients with stable coronary artery disease: results from the Ludwigshafen risk and cardiovascular health study. Clin Chem 60: 530-540. doi: 10.1373/clinchem.2013.209858
![]() |
[117] |
Bayes-Genis A, de Antonio M, Galan A, et al. (2012) Combined use of high-sensitivity ST2 and NTproBNP to improve the prediction of death in heart failure. Eur J Heart Fail 14: 32-38. doi: 10.1093/eurjhf/hfr156
![]() |
[118] | Lippi G, Salvagno GL, Robuschi F, et al. (2014) Influence of dipyridamole stress echocardiography on galectin-3, amino-terminal B-type natriuretic peptide (NT-proBNP) and high-sensitivity troponin T. Acta Cardiol 69: 377-383. |
[119] | Krzeslak A, Lipinska A (2004) Galectin-3 as a multifunctional protein. Cell Mol Biol Lett 9: 305-328. |
[120] |
Sharma UC, Pokharel S, van Brakel TJ, et al. (2004) Galectin-3 marks activated macrophages in failure-prone hypertrophied hearts and contributes to cardiac dysfunction. Circulation 110: 3121-3128. doi: 10.1161/01.CIR.0000147181.65298.4D
![]() |
[121] | Suarez G, Meyerrose G (2014) Heart failure and galectin 3. Ann Transl Med 2: 86. |
[122] |
Chen A, Hou W, Zhang Y, et al. (2015) Prognostic value of serum galectin-3 in patients with heart failure: a meta-analysis. Int J Cardiol 182: 168-170. doi: 10.1016/j.ijcard.2014.12.137
![]() |
[123] |
Winter MP, Wiesbauer F, Alimohammadi A, et al. (2016) Soluble galectin-3 is associated with premature myocardial infarction. Eur J Clin Invest 46: 386-391. doi: 10.1111/eci.12605
![]() |
[124] |
Yancy CW, Jessup M, Bozkurt B, et al. (2013) 2013 ACCF/AHA guideline for the management of heart failure: a report of the American College of Cardiology Foundation/American Heart Association Task Force on practice guidelines. Circulation 128: e240-327. doi: 10.1161/CIR.0b013e31829e8776
![]() |
[125] |
Hogas S, Schiller A, Voroneanu L, et al. (2016) Predictive Value for Galectin 3 and Cardiotrophin 1 in Hemodialysis Patients. Angiology 67: 854-859. doi: 10.1177/0003319715623397
![]() |
[126] |
Gurel OM, Yilmaz H, Celik TH, et al. (2015) Galectin-3 as a new biomarker of diastolic dysfunction in hemodialysis patients. Herz 40: 788-794. doi: 10.1007/s00059-015-4303-6
![]() |
[127] |
de Boer RA, van der Velde AR, Mueller C, et al. (2014) Galectin-3: a modifiable risk factor in heart failure. Cardiovasc Drugs Ther 28: 237-246. doi: 10.1007/s10557-014-6520-2
![]() |
[128] |
Hagstrom E, James SK, Bertilsson M, et al. (2016) Growth differentiation factor-15 level predicts major bleeding and cardiovascular events in patients with acute coronary syndromes: results from the PLATO study. Eur Heart J 37: 1325-1333. doi: 10.1093/eurheartj/ehv491
![]() |
[129] |
Jougasaki M, Rodeheffer RJ, Redfield MM, et al. (1996) Cardiac secretion of adrenomedullin in human heart failure. J Clin Invest 97: 2370-2376. doi: 10.1172/JCI118680
![]() |
[130] |
Nishikimi T, Saito Y, Kitamura K, et al. (1995) Increased plasma levels of adrenomedullin in patients with heart failure. J Am Coll Cardiol 26: 1424-1431. doi: 10.1016/0735-1097(95)00338-X
![]() |
[131] |
Maisel A, Mueller C, Nowak R, et al. (2010) Mid-region pro-hormone markers for diagnosis and prognosis in acute dyspnea: results from the BACH (Biomarkers in Acute Heart Failure) trial. J Am Coll Cardiol 55: 2062-2076. doi: 10.1016/j.jacc.2010.02.025
![]() |
[132] |
Giannopoulos G, Deftereos S, Panagopoulou V, et al. (2013) Copeptin as a biomarker in cardiac disease. Curr Top Med Chem 13: 231-240. doi: 10.2174/15680266113139990088
![]() |
[133] | Reid P, Holen I (2009) Pathophysiological roles of osteoprotegerin (OPG). Eur J Cell Biol 88: 1-17. |
[134] |
Nybo M, Rasmussen LM (2008) The capability of plasma osteoprotegerin as a predictor of cardiovascular disease: a systematic literature review. Eur J Endocrinol 159: 603-608. doi: 10.1530/EJE-08-0554
![]() |
[135] |
Rosenberg M, Zugck C, Nelles M, et al. (2008) Osteopontin, a new prognostic biomarker in patients with chronic heart failure. Circ Heart Fail 1: 43-49. doi: 10.1161/CIRCHEARTFAILURE.107.746172
![]() |
[136] |
Rabin KR, Kamari Y, Avni I, et al. (2005) Adiponectin: linking the metabolic syndrome to its cardiovascular consequences. Expert Rev Cardiovasc Ther 3: 465-471. doi: 10.1586/14779072.3.3.465
![]() |
[137] | Zoccali C, Mallamaci F, Panuccio V, et al. (2003) Adiponectin is markedly increased in patients with nephrotic syndrome and is related to metabolic risk factors. Kidney Int Suppl: S98-102. |
[138] | Zoccali C, Mallamaci F, Tripepi G, et al. (2002) Adiponectin, metabolic risk factors, and cardiovascular events among patients with end-stage renal disease. J Am Soc Nephrol 13: 134-141. |
[139] |
Yan AT, Yan RT, Spinale FG, et al. (2006) Plasma matrix metalloproteinase-9 level is correlated with left ventricular volumes and ejection fraction in patients with heart failure. J Card Fail 12: 514-519. doi: 10.1016/j.cardfail.2006.05.012
![]() |
[140] | Kecebas M, Gullulu S, Sag S, et al. (2014) Serum fetuin-A levels in patients with systolic heart failure. Acta Cardiol 69: 399-405. |
[141] |
Morrow DA, Cannon CP, Rifai N, et al. (2001) Ability of minor elevations of troponins I and T to predict benefit from an early invasive strategy in patients with unstable angina and non-ST elevation myocardial infarction: results from a randomized trial. Jama 286: 2405-2412. doi: 10.1001/jama.286.19.2405
![]() |
[142] |
Sandoval Y, Herzog CA, Love SA, et al. (2016) Prognostic Value of Serial Changes in High-Sensitivity Cardiac Troponin I and T over 3 Months Using Reference Change Values in Hemodialysis Patients. Clin Chem 62: 631-638. doi: 10.1373/clinchem.2015.251835
![]() |