Data governance and Gensini score automatic calculation for coronary angiography with deep-learning-based natural language extraction

Feng Li; Mingfeng Jiang; Hongzeng Xu; Yi Chen; Feng Chen; Wei Nie; Li Wang; Feng Li; Mingfeng Jiang; Hongzeng Xu; Yi Chen; Feng Chen; Wei Nie; Li Wang

doi:10.3934/mbe.2024180

Mathematical Biosciences and Engineering

2024, Volume 21, Issue 3: 4085-4103. doi: 10.3934/mbe.2024180

Previous Article Next Article

Research article Special Issues

Data governance and Gensini score automatic calculation for coronary angiography with deep-learning-based natural language extraction

1.
School of Information and Electronic Engineering, Zhejiang Gongshang University, Hangzhou 310018, China
2.
Department of Cardiology, The People's Hospital of Liaoning Province, Liaoning, Shenyang 110011, China
3.
College of Marine Electrical Engineering, Dalian Maritime University, Dalian 116026, China

Received: 09 December 2023 Revised: 24 January 2024 Accepted: 01 February 2024 Published: 23 February 2024

With the widespread adoption of electronic health records, the amount of stored medical data has been increasing. Clinical data, often in the form of semi-structured or unstructured electronic medical records (EMRs), contains rich patient information. However, due to the use of natural language by physicians when composing these records, the effectiveness of traditional methods such as dictionaries, rule matching, and machine learning in the extraction of information from these unstructured texts falls short of clinical standards. In this paper, a novel deep-learning-based natural language extraction method is proposed to overcome current shortcomings in data governance and Gensini score automatic calculation in coronary angiography. A pre-trained model called bidirectional encoder representation from transformers (BERT) with strong text feature representation capabilities is employed as the feature representation layer. It is combined with bidirectional long short-term memory (BiLSTM) and conditional random field (CRF) models to extract both global and local features from the text. The study included an evaluation of the model on a dataset from a hospital in China and it was compared with another model to validate its practical advantages. Hence, the BiLSTM-CRF model was employed to automatically extract relevant coronary angiogram information from EMR texts. The achieved F1 score was 91.19, which is approximately 0.87 higher than the BERT-BiLSTM-CRF model.

Keywords:

Citation: Feng Li, Mingfeng Jiang, Hongzeng Xu, Yi Chen, Feng Chen, Wei Nie, Li Wang. Data governance and Gensini score automatic calculation for coronary angiography with deep-learning-based natural language extraction[J]. Mathematical Biosciences and Engineering, 2024, 21(3): 4085-4103. doi: 10.3934/mbe.2024180

Related Papers:

[1]	Victor Zhenyu Guo . Almost primes in Piatetski-Shapiro sequences. AIMS Mathematics, 2021, 6(9): 9536-9546. doi: 10.3934/math.2021554
[2]	Yukai Shen . $k$ th powers in a generalization of Piatetski-Shapiro sequences. AIMS Mathematics, 2023, 8(9): 22411-22418. doi: 10.3934/math.20231143
[3]	Jinyun Qi, Zhefeng Xu . Almost primes in generalized Piatetski-Shapiro sequences. AIMS Mathematics, 2022, 7(8): 14154-14162. doi: 10.3934/math.2022780
[4]	Yanbo Song . On two sums related to the Lehmer problem over short intervals. AIMS Mathematics, 2021, 6(11): 11723-11732. doi: 10.3934/math.2021681
[5]	Xiaoqing Zhao, Yuan Yi . High-dimensional Lehmer problem on Beatty sequences. AIMS Mathematics, 2023, 8(6): 13492-13502. doi: 10.3934/math.2023684
[6]	Mingxuan Zhong, Tianping Zhang . Partitions into three generalized D. H. Lehmer numbers. AIMS Mathematics, 2024, 9(2): 4021-4031. doi: 10.3934/math.2024196
[7]	Bingzhou Chen, Jiagui Luo . On the Diophantine equations $x^2-Dy^2=-1$ and $x^2-Dy^2=4$ . AIMS Mathematics, 2019, 4(4): 1170-1180. doi: 10.3934/math.2019.4.1170
[8]	Rui Wang, Jiangtao Peng . On the inverse problems associated with subsequence sums of zero-sum free sequences over finite abelian groups Ⅱ. AIMS Mathematics, 2021, 6(2): 1706-1714. doi: 10.3934/math.2021101
[9]	Jinyan He, Jiagui Luo, Shuanglin Fei . On the exponential Diophantine equation $(a(a-l)m^{2}+1)^{x}+(alm^{2}-1)^{y} = (am)^{z}$ . AIMS Mathematics, 2022, 7(4): 7187-7198. doi: 10.3934/math.2022401
[10]	Baria A. Helmy, Amal S. Hassan, Ahmed K. El-Kholy, Rashad A. R. Bantan, Mohammed Elgarhy . Analysis of information measures using generalized type-Ⅰ hybrid censored data. AIMS Mathematics, 2023, 8(9): 20283-20304. doi: 10.3934/math.20231034

Abstract

1. Introduction

Let $q$ be an integer. For each integer $a$ with

$1 \leqslant a < q,\ \ (a, q) = 1,$

we know that ^[1] there exists one and only one $\bar{a}$ with

$1 \leqslant\bar{a} < q$

such that

$a \bar{a} \equiv 1(q).$

Define

$R(q): = \{a: 1 \leqslant a \leqslant q,(a, q) = 1,2 \nmid a+\bar{a}\},$

$r(q): = \# R(q).$

The work ^[2] posed the problem of investigating a nontrivial estimation for $r(q)$ when $q$ is an odd prime. Zhang ^[3,4] gave several asymptotic formulas for $r(q)$ , one of which is:

$r(q) = \frac{1}{2} \phi(q)+O\left(q^{\frac{1}{2}} d^{2}(q) \log ^{2} q\right),$

where $\phi(q)$ is the Euler function and $d(q)$ is the divisor function. Lu and Yi ^[5] studied a generalization of the Lehmer problem over short intervals. Let $n \geqslant 2$ be a fixed positive integer, $q \geqslant 3$ and $c$ be integers with

$(n c, q) = 1.$

They defined

$\begin{align*} r_n\left(\theta_1, \theta_2, c ; q\right) = \#\left\{(a, b) \in\left[1, \theta_1 q\right] \times\left[1, \theta_2 q\right] \mid a b \equiv c( \text{ mod } q), n \nmid a+b\right\}, \end{align*}$

where $0 < \theta_1, \theta_2 \leqslant 1$ , and obtained

$r_n\left(\theta_1, \theta_2, c ; q\right) = \left(1-\frac{1}{n}\right) \theta_1 \theta_2 \varphi(q)+O\left(q^{1 / 2} \tau^6(q) \log ^2 q\right),$

where the $O$ constant depends only on $n$ . In addition, Xi and Yi ^[6] considered generalized Lehmer problem over short intervals. Han and Liu ^[7] gave an upper bound estimation for another generalization of the Lehmer problem over incomplete interval.

Guo and Yi ^[8] also found the Lehmer problem has good distribution properties on Beatty sequences. For fixed real numbers $\alpha$ and $\beta$ , defined by

$\mathcal{B}_{\alpha, \beta}: = (\lfloor\alpha n+\beta\rfloor)_{n = 1}^{\infty}.$

Beatty sequences are linear sequences. Based on the results obtained, we conjecture the Lehmer problem also has good distribution properties in some non-linear sequences.

The Piatetski-Shapiro sequence is a non-linear sequence, defined by

$\mathbb{N}^c = \left\{\left\lfloor n^c\right\rfloor: n \in \mathbb{N}\right\},$

where $c \in \mathbb{R}$ is non-integer with $c > 1$ and $z \in \mathbb{R}$ . This sequence was first introduced by Piatetski-Shapiro ^[9] to study prime numbers in sequences of the form $\lfloor f(n)\rfloor$ , where $f(n)$ is a polynomial. A positive integer is called square-free if it is a product of distinct primes. The distribution of square-free numbers in the Piatetski-Shapiro sequence has been studied extensively. Stux ^[10] found that, as $x$ tends to infinity,

$\begin{align} \sum\limits_{\substack{n \leq x \\ \left\lfloor n^c\right\rfloor \text { is square-free }}} 1 = \left(\frac{6}{\pi^2}+o(1)\right) x, \quad \text { for }\ 1 < c < \frac{4}{3} . \end{align}$

(1.1)

In 1978, Rieger ^[11] improved the range to $1 < c < 3 / 2$ and obtained

$\frac{6 x}{\pi^2}+O\left(x^{(2 c+1) / 4+\varepsilon}\right), \quad \text { for }\ 1 < c < \frac{3}{2} \text {. }$

Considering the results obtained, we develop this problem by investigating

$R(c;q): = \mathop{\sum\limits_{\;\;n \in \mathbb{N}^c \cap R(q)\\n \text { is square-free } }} 1$

and range of $c$ when $q$ tends to infinity. By methods of exponential sum and Kloosterman sums and fairly detailed calculations, we get the following result, which is significant for understanding the distribution properties of the Lehmer problem.

Theorem 1.1. Let $q$ be an odd integer and large enough,

$\gamma : = 1/c \; \; \; \mathit{\text{and}}\; \; \; c \in (1, \frac{4}{3}),$

we obtain

$\begin{align} R(c;q) = &\frac{3}{\pi^{2}}\prod\limits_{p\mid q}(1+p^{-1})^{-1}q^{\gamma} \\&+O\left( \sum\limits_{p\mid q}(1-p^{-\frac{1}{2}}) ^{-1} q^{\gamma-\frac{1}{2}}\right)+O\left(q^{\frac{7}{13}\gamma+\frac{4}{13}}\prod\limits_{p\mid q}(1-p^{-\frac{1}{2}})^{-1}\log q\right)\\ &+O\left( q^{\frac{3}{4}}d^{3}(q)\log q \right)+O\left(q^{ \gamma -\frac{1}{6}} d^{2}(q)\log^3 q\right), \end{align}$

where the $O$ constant only depends on c.

This paper consists of three main sections. Introduction covers the origins and developments of the Lehmer problem, along with several interesting results. It also presents relevant findings related to the Piatetski-Shapiro sequences. The second section includes some definitions and lemmas throughout the paper. The third section outlines the calculation process, where we use additive characteristics to convert the congruence equations into exponential sum problems. We then employ the Kloosterman sums and exponential sums methods to derive an interesting asymptotic formula.

2. Preliminary lemmas

To complete the proof of the theorem, we need the following several definitions and lemmas.

In this paper, we denote by $\lfloor t\rfloor$ and $\{t\}$ the integral part and the fractional part of $t$ , respectively. As is customary, we put

$\mathbf{e}(t): = e^{2 \pi \mathbf{i} t} \quad \text {and} \quad\{t\}: = t-\lfloor t\rfloor.$

The notation $\|t\|$ is used to denote the distance from the real number $t$ to the nearest integer; that is,

$\|t\|: = \min\limits_{n \in \mathbb{Z}}|t-n|.$

And $\sum^{\prime}$ indicates that the variable summed over takes values coprime to the number $q$ . Throughout the paper, $\varepsilon$ always denotes an arbitrarily small positive constant, which may not be the same at different occurrences; the implied constants in symbols $O, \ll$ , and $\gg$ may depend (where obvious) on the parameters $c \text{ and } \varepsilon$ , but are absolute otherwise. For given functions $F$ and $G$ , the notations

$F \ll G,\ \ \ G \gg F\ \ \ \text{and}\; \; \; F = O(G)$

are all equivalent to the statement that the inequality

$|F| \leqslant \mathcal{C}|G|$

holds with some constant $\mathcal{C} > 0$ .

Lemma 2.1. Let $\mathbf{1}_{c}(m)$ denote the characteristic function of numbers in a Piatetski-Shapiro sequence, then

$\mathbf{1}_{c}(m) = \gamma m^{\gamma-1} +O(m^{\gamma-2}) + \psi \left(-(m+1)^{\gamma}\right) -\psi(-m^{\gamma})\nonumber ,$

where

$\psi(t) = t- \lfloor t \rfloor-\frac{1}{2} \ \ \ \mathit{\text{and}}\; \; \; \gamma = 1/c.$

Proof. Note that an integer $m$ has the form

$m = \lfloor n^c\rfloor$

for some integer $n$ if and only if

$m \leqslant n^{c} < m+1,\quad -(m+1)^{\gamma} < - n \leqslant -m^{\gamma}.$

$\begin{align} \mathbf{1}_{c}(m)& = \left\lfloor-m^{\gamma}\right\rfloor-\left\lfloor-(m+1)^{\gamma}\right\rfloor \\ & = -m^{\gamma} - \psi (-m^{\gamma}) +(m+1)^{\gamma}+\psi(-(m+1)^{\gamma})\\ & = \gamma m^{\gamma-1} +O(m^{\gamma-2})+ \psi\left (-(m+1)^{\gamma}\right) -\psi(-m^{\gamma}). \end{align}$

Thie completes the proof. □

Lemma 2.2. Let $H \geqslant 1$ be an integer, $a_h, b_h$ be real numbers, we have

$\left|\psi(t)-\sum\limits_{0 < |h| \leqslant H} a_h \mathbf{e}(t h)\right| \leqslant \sum\limits_{|h| \leqslant H} b_h \mathbf{e}(t h), \quad a_h \ll \frac{1}{|h|}, \quad b_h \ll \frac{1}{H}.$

Proof. In 1985, Vaaler showed how Beurling's function could be used to construct a trigonometric polynomial approximation to $\psi (x).$ For each positive integer $N$ , Vaaler's construction yields a trigonometric polynomial $\psi^*$ of degree $N$ which satisfies

$|\psi^*(x)-\psi(x)|\leq\frac1{2N+2}\sum\limits_{|n|\leq N}\left(1-\frac{|n|}{N+1}\right)\mathbf{e}(nx),$

where

$\begin{align*} \psi^*(x)& = -\mathop{\sum}_{1\leq|n|\leq N}(2\pi in)^{-1}\hat{J}_{N+1}(n)\mathbf{e}(nx), \\ H(z)& = \frac{\sin^2\pi z}{\pi^2}\left\{\sum\limits_{n = -\infty}^\infty\frac{\mathrm{sgn\; }(n)}{(z-n)^2}+\frac2z\right\},\quad J(z) = \frac12H^{\prime}(z),\\ H_N(z)& = \frac{\sin^2\pi z}{\pi^2}\left\{\sum\limits_{|n|\leq N}\frac{\mathrm{sgn\; }(n)}{(z-n)^2}+\frac2z\right\}, \quad J_N(z) = \frac12H_N^{\prime}(z), \end{align*}$

and $\mathrm{sgn\; }(n)$ is the sign of $n$ . The Fourier transform $\hat{J}(t)$ satisfies

$\begin{align} \hat{J}(t) = \begin{cases} 1 , & t = 0 ;\\ \pi t(1-|t|)\cot\pi t+|t|,& 0 < |t| < 1;\\ 0,& t\geq1.\end{cases} \end{align}$

To be short, we denote

$\begin{align*} a_h& = -(2\pi ih)^{-1}\hat{J}_{H+1}(h) \ll \frac{1}{|h|},\\ b_h& = \frac1{2H+2}\left(1-\frac{|h|}{H+1}\right)\ll \frac{1}{H}. \end{align*}$

There are more details in Appendix Theorem A.6. of ^[12]. □

Lemma 2.3. Denote

$\mathbf{Kl}(m,n;q) = \mathop{\sum\limits_{a = 1}^q \sum\limits_{b = 1}^q}_{a b \equiv 1( \bmod q)}\mathbf{e}\left(\frac{m a+n b}{q}\right),$

then

$\mathbf{Kl}(m,n;q) \ll(m, n, q)^{\frac{1}{2}} q^{\frac{1}{2}} d(q),$

where $(m, n, q)$ is the greatest common divisor of $m, n$ and $q$ and $d(q)$ is the number of positive divisors of $q$ .

Proof. The proof is given in ^[13]. □

Lemma 2.4. (Korobov ^[14]) Let $\alpha$ be a real number, $Q$ be an integer, and $P$ be a positive integer, then

$\left|\sum\limits_{x = Q+1}^{Q+P} \mathbf{e}(\alpha x)\right| \leqslant \min \left(P, \frac{1}{2\|\alpha\|}\right) .$

Lemma 2.5. (Karatsuba ^[15]) For any number $b$ , $U < 0$ , $K \geqslant 1$ , let

$a = \frac{s}{r}+\frac{\theta}{r^{2}}, \quad(r, s) = 1,\ \ \ r \geqslant 1,\ \ \ |\theta| \leqslant 1,$

then

$\sum\limits_{k \leqslant K} \min \left(U, \frac{1}{\|a k+b\|}\right) \ll\left(\frac{K}{r}+1\right)(U+r \log r).$

Lemma 2.6. Suppose $f$ is continuously differentiable, $f^{\prime}(n)$ is monotonic, and

$\left\|f^{\prime}(n)\right\| \geqslant \lambda_1 > 0$

on $I$ , then

$\sum\limits_{n \in I} \mathbf{e}\left(f(n)\right) \ll \lambda_1^{-1} .$

Proof. See [12, Theorem 2.1]. □

Lemma 2.7. Let $k$ be a positive integer, $k \geqslant 2$ . Suppose that $f(n)$ is a real-valued function with $k$ continuous derivatives on $[N, 2N],$ Further suppose that

$0 < F \leqslant f^{(k)}(n) \leqslant h F.$

Then

$\left|\sum\limits_{N < x \leqslant 2N} \mathbf{e}(f(n)) \right| \ll F^{\kappa} N^\lambda + F^{-1},$

where the implied constant is absolute.

Proof. See [12, Chapter 3]. □

3. Proof of theorem

By the definition of Mobius function

$\begin{align} \mu(n) = \begin{cases} (-1)^{\omega(n)} , & \forall p|n, p^2 \nmid n ,\\ 0,& \quad \exists p^2 |n ,\end{cases} \end{align}$

it is clear that $n$ is square-free if and only if

$\mu^{2}(n) = 1,$

where $\omega(n)$ is the number of prime divisor of $n$ . So

$\begin{align} R(c; q)& = \frac{1}{2} \mathop{{\sum}^{\prime}}_{n = 1}^{q} \left(1-(-1)^{n+\bar{n}}\right)\mu^{2}(n) \mathbf{1}_{c}(n)\\ & = \frac{1}{2}(R_{1}-R_{2}), \end{align}$

(3.1)

where

$\begin{align} R_{1} = \mathop{{\sum}^{\prime}}_{n = 1}^{q} \mu^{2}(n) \mathbf{1}_{c}(n) \end{align}$

and

$\begin{align} R_{2} = \mathop{{\sum}^{\prime}}_{n = 1}^{q} (-1)^{n+\bar{n}}\mu^{2}(n) \mathbf{1}_{c}(n). \end{align}$

3.1. Estimation of $R_{1}$

From Lemma 2.1, we have

$\begin{align} R_{1} & = \mathop{{\sum}^{\prime}}_{n = 1}^{q} \mu^{2}(n) \mathbf{1}_{c}(n) \\ & = \mathop{{\sum}^{\prime}}_{n = 1}^{q} \mu^{2}(n)\left( \gamma n^{\gamma-1} +O(n^{\gamma-2}) + \psi\left (-(n+1)^{\gamma}\right) -\psi(-n^{\gamma}) \right) \\ & = R_{11}+R_{12} , \end{align}$

(3.2)

where

$\begin{align} R_{11}&: = \mathop{{\sum}^{\prime}}_{n = 1}^{q} \mu^{2}(n)\left( \gamma n^{\gamma-1} +O(n^{\gamma-2}) \right)\\ & = \mathop{{\sum}^{\prime}}_{n = 1}^{q} \mu^{2}(n)\gamma n^{\gamma-1}+ O\left(\mathop{{\sum}^{\prime}}_{n = 1}^{q} \mu^{2}(n)n^{\gamma-2}\right) . \end{align}$

Let

$\mathcal{D} = \{d: p|d \Rightarrow p| q\}$

and $\lambda(n)$ is Liouville function. When $n \in R(q)$ ,

$\begin{align} \mu^{2}(n) = \begin{cases} \sum\limits_{dm = n,d\in \mathcal{D}} \lambda(d) \mu^{2}(m), &(n,q) = 1 ,\\ 0,& (n,q) > 1 .\end{cases} \end{align}$

(3.3)

We just consider the first term of $R_{11}$ . Applying Euler summuation ^[1],

$\begin{align} \mathop{{\sum}^{\prime}}_{n = 1}^{q} \mu^{2}(n)\gamma n^{\gamma-1} & = \mathop{\sum}_{n = 1}^{q} \mathop{\sum\limits_{dm = n }}_{d\in \mathcal{D}} \lambda(d) \mu^{2}(m) \gamma(dm)^{\gamma-1}\\ & = \sum\limits_{d\in \mathcal{D} } \lambda(d)d^{\gamma-1} \sum\limits_{m\leqslant\frac{q}{d}}\mu^{2}(m) \gamma m^{\gamma-1}\\ & = \sum\limits_{d\in \mathcal{D}} \lambda(d)d^{\gamma-1} \sum\limits_{m\leqslant\frac{q}{d}}\left(\sum\limits_{l^2\mid m}\mu (l)\right) \gamma m^{\gamma-1}\\ & = \sum\limits_{d\in \mathcal{D}} \lambda(d)d^{\gamma-1} \sum\limits_{l\leqslant(\frac{q}{d})^{\frac{1}{2}}}\mu (l)l^{2\gamma-2} \sum\limits_{m\leqslant\frac{q}{dl^2}}\gamma m^{\gamma-1}\\ & = \sum\limits_{d\in \mathcal{D}} \lambda(d)d^{\gamma-1} \sum\limits_{l\leqslant(\frac{q}{d})^{\frac{1}{2}}}\mu (l)l^{2\gamma-2}\left((\frac{q}{dl^2})^{\gamma}+O\left((\frac{q}{dl^2})^{\gamma-1}\right)\right) \\ & = q^{\gamma}\sum\limits_{d\in \mathcal{D}} \lambda(d)d^{-1} \sum\limits_{l\leqslant(\frac{q}{d})^{\frac{1}{2}}}\mu (l)l^{ -2} +O\left(\sum\limits_{d\in \mathcal{D}} \sum\limits_{l\leqslant(\frac{q}{d})^{\frac{1}{2}}} q^{\gamma-1}\right)\\ & = q^{\gamma}\sum\limits_{d\in \mathcal{D}} \lambda(d)d^{-1} \left(\sum\limits_{l}\mu (l)l^{ -2}+O\left((\frac{q}{d})^{-\frac{1}{2}}\right)\right) +O\left( \prod\limits_{p\mid q}(1-p^{-\frac{1}{2}}) ^{-1} q^{\gamma-\frac{1}{2}}\right)\\ & = \frac{6}{\pi^{2}}\prod\limits_{p\mid q}(1+p^{-1})^{-1}q^{\gamma} +O\left( \prod\limits_{p\mid q}(1-p^{-\frac{1}{2}}) ^{-1} q^{\gamma-\frac{1}{2}}\right), \end{align}$

thus

$\begin{align} R_{11} = \frac{6}{\pi^{2}}\prod\limits_{p\mid q}(1+p^{-1})^{-1}q^{\gamma} +O\left( \prod\limits_{p\mid q}(1-p^{-\frac{1}{2}}) ^{-1} q^{\gamma-\frac{1}{2}}\right). \end{align}$

(3.4)

For $R_{12}$ , by Lemma 2.2, we have

$\begin{align} R_{12}&: = \mathop{{\sum}^{\prime}}_{n = 1}^{q} \mu^{2}(n) \left(\psi\left(-(n+1)^{\gamma}\right)-\psi(-n^{\gamma})\right)\\ & = R_{121}+O(R_{122}), \end{align}$

(3.5)

where

$\begin{align} R_{121}: = \mathop{{\sum}^{\prime}}_{n = 1}^{q} \mu^{2}(n) \left(\sum\limits_{0 < |h|\leqslant H}a_{h}\left(\mathbf{e}\left(-(n+1)^{\gamma}h\right)-\mathbf{e}\left(-n^{\gamma}h\right)\right)\right) \end{align}$

and

$\begin{align} R_{122}: = \mathop{{\sum}^{\prime}}_{n = 1}^{q} \mu^{2}(n) \left(\sum\limits_{|h|\leqslant H}b_{h}\left(\mathbf{e}\left(-(n+1)^{\gamma}h\right)+\mathbf{e}\left(-n^{\gamma}h\right)\right)\right ). \end{align}$

Define

$f(t) = \mathbf{e}\left(\left((dt)^{\gamma}-(dt+1)^{\gamma}\right)h\right)-1,$

then

$\begin{align} f(t) &\ll\ |h| (dt)^{\gamma-1}, \\ \frac{\partial f(t)}{\partial t} &\ll |h| d^{\gamma-1}t^{\gamma-2}. \end{align}$

By Lemma 2.2 and Eq (3.3),

$\begin{align} R_{121} = &\sum\limits_{0 < |h|\leqslant H}a_{h} \sum\limits_{d\in \mathcal{D}} \lambda(d) \left(\sum\limits_{1 < m \leqslant \frac{q}{d}} \mu^{2}(m) \mathbf{e}\left(- (dm)^{\gamma}h\right) f(m)\right) \\ \ll& \sum\limits_{0 < |h|\leqslant H}|h|^{-1} \sum\limits_{d\in \mathcal{D}} \left|\int_{0}^{ \frac{q}{d}} f(t) \mathrm{d}\left(\sum\limits_{1 < m \leqslant t} \mu^{2}(m) \mathbf{e}\left(- (dm)^{\gamma}h\right)\right) \right|\\ \ll& \sum\limits_{0 < |h|\leqslant H}|h|^{-1} \sum\limits_{d\in \mathcal{D}} \left| f(\frac{q}{d}) \left(\sum\limits_{1 < m \leqslant \frac{q}{d}} \mu^{2}(m) \mathbf{e}\left(- (dm)^{\gamma}h\right)\right) \right|\\ &+\sum\limits_{0 < |h|\leqslant H}|h|^{-1} \sum\limits_{d\in \mathcal{D}} \left|\int_{0}^{ \frac{q}{d}} \frac{\partial f(t)}{\partial t} \sum\limits_{1 < m \leqslant t} \mu^{2}(m) \mathbf{e}\left(- (dm)^{\gamma}h\right)\mathrm{d}t \right|. \end{align}$

Let

$(\kappa,\lambda) = (\frac{1}{6},\frac{2}{3})$

be an exponential pair. Applying Lemma 2.7, it's easy to see

$\begin{align} \sum\limits_{1 < m \leqslant t} \mu^{2}(m) \mathbf{e}\left(- (dm)^{\gamma}h\right) = &\sum\limits_{m\leqslant t} \left(\sum\limits_{l^{2}\mid m}\mu (l)\right) \mathbf{e}\left(- (dm)^{\gamma}h\right) \\ \ll& \sum\limits_{l\leqslant t^{\frac{1}{2}}}\left| \sum\limits_{m\leqslant \frac{t}{l^{2}}}\mathbf{e}\left(- (dl^2m)^{\gamma}h\right)\right| \\ \ll&\sum\limits_{l\leqslant t^{\frac{1}{2}}} \log q \left(\left((dl^2)^{\gamma}|h|(\frac{t}{l^{2}})^{\gamma-1}\right)^{\frac{1}{6}}(\frac{t}{l^{2}})^{\frac{2}{3} }+\left((dl^2)^{\gamma}|h|(\frac{t}{l^{2}})^{\gamma-1}\right)^{-1} \right)\\ \ll& \log q \sum\limits_{l\leqslant t^{\frac{1}{2}}}\left((d^{\gamma}|h|)^{\frac{1}{6}}{t}^{\frac{1}{6}\gamma+\frac{1}{2}}l^{-1}+(d^{\gamma}|h|)^{-1}t^{1-\gamma }l^{-2}\right)\\ \ll& (d^{\gamma}|h|)^{\frac{1}{6}}{t}^{\frac{1}{6}\gamma+\frac{1}{2}} \log^2 q+(d^{\gamma}|h|)^{-1}t^{1-\gamma }\log q \\ \ll&(d^{\gamma}|h|)^{\frac{1}{6}}{t}^{\frac{1}{6}\gamma+\frac{1}{2}} \log^2 q, \end{align}$

thus

$\begin{align} R_{121}\ll& \sum\limits_{0 < |h|\leqslant H}|h|^{-1} \sum\limits_{d\in \mathcal{D}} |h|q^{\gamma-1}(d^{\gamma}|h|)^{\frac{1}{6}}{(\frac{q}{d})}^{\frac{1}{6}\gamma+\frac{1}{2}} \log^2 q \\ &+\sum\limits_{0 < |h|\leqslant H}|h|^{-1} \sum\limits_{d\in \mathcal{D}} \int_{0}^{ \frac{q}{d}} | h| d^{\gamma-1}t^{\gamma-2}(d^{\gamma}|h|)^{\frac{1}{6}}{t}^{\frac{1}{6}\gamma+\frac{1}{2}} \log^2 q \mathrm{d}t \\ \ll& \sum\limits_{0 < |h|\leqslant H}|h|^{\frac{1}{6}}\sum\limits_{d\in \mathcal{D}}d^{-\frac{1}{2}}q^{\frac{7}{6}\gamma-\frac{1}{2}} \log^2 q \\ &+\sum\limits_{0 < |h|\leqslant H}|h|^{\frac{1}{6}} \sum\limits_{d\in \mathcal{D}}d^{\frac{7}{6}\gamma-1}\log^2 q\int_{0}^{ \frac{q}{d}} t^{\frac{7}{6}\gamma-\frac{3}{2}} \mathrm{d}t\\ \ll& H^{\frac{7}{6}}\prod\limits_{p\mid q}(1-p^{-\frac{1}{2}})^{-1}q^{\frac{7}{6}\gamma-\frac{1}{2}}\log^2 q. \end{align}$

(3.6)

For $R_{122}$ , the contribution from $h \neq0$ can be bounded by similar methods of Eq (3.6). Taking

$H = q^{\frac{9}{13}-\frac{7}{13}\gamma} \geqslant 1,$

we obtain

$\begin{align} R_{122} & = b_{0} \mathop{{\sum}^{\prime}}_{n = 1}^{q} \mu^{2}(n)+\sum\limits_{0 < |h|\leqslant H}b_{h} \mathop{{\sum}^{\prime}}_{n = 1}^{q} \mu^{2}(n) \left(e\left(-(n+1)^{\gamma}h\right)+e\left(-n^{\gamma}h\right)\right) \\ &\ll H^{-1}q + H^{\frac{7}{6}}\prod\limits_{p\mid q}(1-p^{-\frac{1}{2}})^{-1}q^{\frac{7}{6}\gamma-\frac{1}{2}}\log q \\ &\ll q^{\frac{7}{13}\gamma+\frac{4}{13}}\prod\limits_{p\mid q}(1-p^{-\frac{1}{2}})^{-1}\log^2 q. \end{align}$

(3.7)

It follows from Eqs ( $3.5$ )–( $3.7$ ),

$\begin{align} R_{12} \ll q^{\frac{7}{13}\gamma+\frac{4}{13}}\prod\limits_{p\mid q}(1-p^{-\frac{1}{2}})^{-1}\log^2 q . \end{align}$

Hence

$\begin{align} R_{1} = \frac{6}{\pi^{2}}\prod\limits_{p\mid q}(1+p^{-1})^{-1}q^{\gamma} +O\left( \sum\limits_{p\mid q}(1-p^{-\frac{1}{2}}) ^{-1} q^{\gamma-\frac{1}{2}}\right)+O\left(q^{\frac{7}{13}\gamma+\frac{4}{13}}\prod\limits_{p\mid q}(1-p^{-\frac{1}{2}})^{-1}\log^2 q\right). \end{align}$

(3.8)

3.2. Estimation of $R_{2}$

Similarly,

$\begin{align} R_{2} = & \mathop{{\sum}^{\prime}}_{n = 1}^{q} (-1)^{n+\bar{n}} \mu^{2}(n) \mathbf{1}_{c}(n) \\ = & RP_{21}+RP_{22} , \end{align}$

(3.9)

where

$\begin{align} R_{21} = \mathop{{\sum}^{\prime}}_{n = 1}^{q}(-1)^{n+\bar{n}} \mu^{2}(n)\left( \gamma n^{\gamma-1} +O(n^{\gamma-2}) \right) \end{align}$

and

$\begin{align} R_{22} = \mathop{{\sum}^{\prime}}_{n = 1}^{q}(-1)^{n+\bar{n}} \mu^{2}(n)\left(\psi\left(-(n+1)^{\gamma}\right)-\psi(-n^{\gamma})\right). \end{align}$

We also just consider the first term of $R_{21}$ .

$\begin{align} \mathop{{\sum}^{\prime}}_{n = 1}^{q}(-1)^{n+\bar{n}} \mu^{2}(n) \gamma n^{\gamma-1} = &\mathop{{\sum}^{\prime}}_{n = 1}^{q} (-1)^{n+\bar{n}}\left( \sum\limits_{d^{2}\mid n} \mu(d) \right) \gamma n^{\gamma-1} \\ = &\mathop{{\sum}^{\prime}}_{n = 1}^{q} (-1)^{n+\bar{n}} \mathop{\sum\limits_{d^{2}\mid n}}_{d \leqslant q^{\frac{1}{4} }} \mu(d) \gamma n^{\gamma-1}+\mathop{{\sum}^{\prime}}_{n = 1}^{q} (-1)^{n+\bar{n}} \mathop{\sum\limits_{d^{2}\mid n}}_{q^{\frac{1}{4} } < d \leqslant q^{\frac{1}{2}} } \mu(d) \gamma n^{\gamma-1}. \end{align}$

(3.10)

It is easy to see

$\begin{align} \mathop{{\sum}^{\prime}}_{n = 1}^{q} (-1)^{n+\bar{n}} \mathop{\sum\limits_{d^{2}\mid n}}_{q^{\frac{1}{4} } < d \leqslant q^{\frac{1}{2}} } \mu(d) \gamma n^{\gamma-1} \ll \mathop{{\sum}^{\prime}}_{n = 1}^{q} \mathop{\sum\limits_{d^{2}\mid n}}_{q^{\frac{1}{4} } < d \leqslant q^{\frac{1}{2}} } \gamma n^{\gamma-1} \ll q^{\gamma-\frac{1}{4} }. \end{align}$

(3.11)

Since for integers $m$ and $a$ , one has

$\frac{1}{q}\sum\limits_{s = 1}^{q}\mathbf{e}(\frac{s(m-a)}{q}) = \begin{cases} 1, &m \equiv a (\text{ mod } q);\\ 0, &m \not \equiv a (\text{ mod } q).\end{cases}$

This gives

$\begin{align} \mathop{{\sum}^{\prime}}_{n = 1}^{q} (-1)^{n+\bar{n}} \mathop{\sum\limits_{d^{2}\mid n}}_{d \leqslant q^{\frac{1}{4} }} \mu(d) \gamma n^{\gamma-1} = &\mathop{\sum\limits_{n = 1}^{q} \sum\limits_{m = 1}^{q} }_{nm \equiv 1 (\bmod q)} (-1)^{n+m} \mathop{\sum\limits_{d^{2}\mid n}}_{d \leqslant q^{\frac{1}{4} }} \mu(d) \gamma n^{\gamma-1} \mathop{\sum\limits_{a = 1}^{q-1}}_{a = m } \mathop{\sum\limits_{b = 1}^{q-1}}_{b = n } 1\\ = &\mathop{\sum\limits_{n = 1}^{q} \sum\limits_{m = 1}^{q} }_{nm \equiv 1 (\bmod q)} \mathop{\sum\limits_{a = 1}^{q-1}}_{a = m } \mathop{\sum\limits_{b = 1}^{q-1}}_{b = n } (-1)^{a+b} \mathop{\sum\limits_{d^{2}\mid b}}_{d \leqslant q^{\frac{1}{4} }} \mu(d) \gamma b^{\gamma-1} \\ = & \mathop{\sum\limits_{n = 1}^{q} \sum\limits_{m = 1}^{q} }_{nm \equiv 1 (\bmod q)} \mathop{\sum\limits_{a = 1}^{q-1}}_{a \equiv m (\bmod q) } \mathop{\sum\limits_{b = 1}^{q-1}}_{b \equiv n (\bmod q) }(-1)^{a+b}\mathop{\sum\limits_{d^{2}\mid b}}_{d \leqslant q^{\frac{1}{4} }} \mu(d) \gamma b^{\gamma-1} \\ = & \mathop{\sum\limits_{n = 1}^{q} \sum\limits_{m = 1}^{q} }_{nm \equiv 1 (\bmod q)} \sum\limits_{a = 1}^{q-1}\sum\limits_{b = 1}^{q-1}(-1)^{a+b}\mathop{\sum\limits_{d^{2}\mid b}}_{d \leqslant q^{\frac{1}{4} }} \mu(d) \gamma b^{\gamma-1} \\ &\times\left(\frac{1}{q}\sum\limits_{s = 1}^{q}\mathbf{e}(\frac{s(m-a)}{q}) \right)\left(\frac{1}{q}\sum\limits_{t = 1}^{q}\mathbf{e}(\frac{t(n-b)}{q})\right) \\ = &\frac{1}{q^{2}}\sum\limits_{s = 1}^{q}\sum\limits_{t = 1}^{q} \left( \mathop{\sum}_{nm \equiv 1 (\bmod q)} \mathbf{e}(\frac{sm+tn}{q}) \right)\\ &\times\left(\sum\limits_{a = 1}^{q-1}(-1)^{a}\mathbf{e}(-\frac{sa}{q})\right)\left(\sum\limits_{b = 1}^{q-1}(-1)^{b}\mathbf{e}(-\frac{tb}{q})\mathop{\sum\limits_{d^{2}\mid b}}_{d \leqslant q^{\frac{1}{4} }} \mu(d) \gamma b^{\gamma-1} \right ). \end{align}$

(3.12)

From Lemma 2.3,

$\begin{align} \mathop{\sum}_{nm \equiv 1 (\bmod q)} \mathbf{e}(\frac{sm+tn}{q}) = \mathbf{Kl}(s,t;q) \ll(s, t, q)^{\frac{1}{2}} q^{\frac{1}{2}} d(q) . \end{align}$

(3.13)

Note the estimate

$\begin{align} \left|\sum\limits_{a = 1}^{q-1}(-1)^{a}\mathbf{e}(-\frac{sa}{q}) \right| \ll\frac{1}{\left|e(\frac{1}{2}-\frac{s}{q})-1\right|}\ll\frac{1}{|\cos \frac{s}{q}\pi |} \end{align}$

(3.14)

holds. By Abel summation and Lemma 2.4, we have

$\begin{align} \sum\limits_{b = 1}^{q-1}(-1)^{b}\mathbf{e}(-\frac{tb}{q})\mathop{\sum\limits_{d^{2}\mid b}}_{d \leqslant q^{\frac{1}{4}}} \mu(d) \gamma b^{\gamma-1} = &\sum\limits_{d \leqslant q^{\frac{1}{4}}} \mu(d)\sum\limits_{b = 1}^{\lfloor\frac{q-1}{d^2}\rfloor}(-1)^{d^{2}b}\mathbf{e}(-\frac{td^{2}b}{q}) \gamma (d^2b)^{\gamma-1} \\ \ll&\sum\limits_{d \leqslant q^{\frac{1}{4}}} d^{2(\gamma-1)}\left|\sum\limits_{b = 1}^{\lfloor\frac{q-1}{d^2}\rfloor}(-1)^{d^{2}b}\mathbf{e}(-\frac{td^{2}b}{q}) \gamma b^{\gamma-1}\right| \\ \ll& \sum\limits_{d \leqslant q^{\frac{1}{4}}} d^{2(\gamma-1)}\gamma(\frac{q}{d^2})^{\gamma-1} \max\limits_{1 \leqslant \beta \leqslant \lfloor\frac{q-1}{d^2}\rfloor }\left| \sum\limits_{b = 1}^{\beta }(-1)^{d^{2}b}\mathbf{e}(-\frac{td^{2}b}{q}) \right| \\ \ll& \mathop{\sum\limits_{d \leqslant q^{\frac{1}{4}}}}_{2 \mid d}\gamma q^{ \gamma-1 }\max\limits_{1 \leqslant \beta \leqslant \lfloor\frac{q-1}{d^2}\rfloor }\left| \sum\limits_{b = 1}^{ \beta} \mathbf{e}(-\frac{td^{2}b}{q}) \right| \\ & +\mathop{\sum\limits_{d \leqslant q^{\frac{1}{4}}}}_{2 \nmid d} \gamma q^{ \gamma-1 }\max\limits_{1 \leqslant \beta \leqslant \lfloor\frac{q-1}{d^2}\rfloor }\left| \sum\limits_{b = 1}^{\beta }(-1)^{ b}\mathbf{e}(-\frac{td^{2}b}{q}) \right| \\ \ll& \mathop{\sum\limits_{d \leqslant q^{\frac{1}{4} }}}_{2 \mid d}q^{ \gamma-1 }\min\left( \lfloor\frac{q-1}{d^2}\rfloor,\frac{1}{2\|\frac{d^{2}}{q}t \|} \right)+\mathop{\sum\limits_{d \leqslant q^{\frac{1}{4} }}}_{2 \nmid d}q^{ \gamma-1 }\min\left( \lfloor\frac{q-1}{d^2}\rfloor,\frac{1}{2\|\frac{1}{2}-\frac{d^{2}}{q}t \|} \right) . \end{align}$

(3.15)

To be short, combining Eqs (3.13)–(3.15), we denote

$\begin{align} R_{211}: = &q^{-2}\sum\limits_{s = 1}^{q}\sum\limits_{t = 1}^{q}(s,t,q)^{\frac{1}{2}}d(q)q^{\frac{1}{2}}\frac{1}{|\cos\frac{s}{q}\pi|} \mathop{\sum\limits_{d \leqslant q^{\frac{1}{4} }}}_{2 \mid d}q^{ \gamma-1 }\min\left( \lfloor\frac{q-1}{d^2}\rfloor,\frac{1}{2\|\frac{d^{2}}{q}t \|} \right)\\ \ll& q^{ \gamma-3}\sum\limits_{s = 1}^{q}\sum\limits_{t = 1}^{q}(s,t,q)^{\frac{1}{2}}d(q)q^{\frac{1}{2}}\frac{1}{|\cos\frac{s}{q}\pi|} \sum\limits_{d \leqslant q^{\frac{1}{4} }} \min\left( \frac{q-1}{d^2},\frac{1}{2\|\frac{d^{2}}{q}t \|} \right)\\ \ll& q^{\gamma-3}\sum\limits_{u\mid q}u^{\frac{1}{2}} d(q)q^{\frac{1}{2}}\sum\limits_{s = 1}^{\frac{q}{u}}\frac{1}{|\cos\frac{su}{q}\pi|} \sum\limits_{d \leqslant q^{\frac{1}{4} }} \sum\limits_{t = 1}^{\frac{q}{u}}\min\left( \frac{q-1}{d^2},\frac{1}{2\|\frac{d^{2}}{\frac{q}{u}}t \|} \right) \end{align}$

and

$R_{212}: = q^{-2}\sum\limits_{s = 1}^{q}\sum\limits_{t = 1}^{q}(s,t,q)^{\frac{1}{2}}d(q)q^{\frac{1}{2}}\frac{1}{|\cos\frac{s}{q}\pi|} \mathop{\sum\limits_{d \leqslant q^{\frac{1}{4} }}}_{2 \nmid d}q^{ \gamma-1 }\min\left( \lfloor\frac{q-1}{d^2}\rfloor,\frac{1}{2\|\frac{1}{2}-\frac{d^{2}}{q}t \|} \right).\nonumber$

Let

$(d^{2},\frac{q}{u}) = r,\ \ \ (\frac{d^{2}}{r},\frac{q}{ur}) = 1,$

making use of Lemma 2.5, we have

$\sum\limits_{t = 1}^{\frac{q}{u}}\min\left( \frac{q-1}{d^2},\frac{1}{2\|\frac{d^{2}}{\frac{q}{u}}t \|} \right)\ll (\frac{\frac{q}{u}}{\frac{q}{ur}}+1)(\frac{q-1}{d^2}+\frac{q}{ur}\log q) \ll \frac{qr}{d^2}+ \frac{q }{u} \log q.$

Insert it to $R_{211}$ , then

$\begin{align} R_{211}\ll& q^{\gamma-3}\sum\limits_{u\mid q}u^{\frac{1}{2}} d(q)q^{\frac{1}{2}}\sum\limits_{s = 1}^{\frac{q}{u}}\frac{1}{|1-2 \frac{su}{q} |} \sum\limits_{d \leqslant q^{\frac{1}{4} }}\sum\limits_{(d^{2},\frac{q}{u}) = r}( \frac{qr}{d^2}+ \frac{q }{u} \log q)\\ \ll&q^{\gamma-3}\sum\limits_{u\mid q}u^{\frac{1}{2}} d(q)q^{\frac{1}{2}}\frac{q}{u}\log q\sum\limits_{d \leqslant q^{\frac{1}{4} }}\mathop{\sum\limits_{r |d^{2}}}_{r|\frac{q}{u} }( \frac{qr}{d^2}+ \frac{q }{u} \log q) \\ \ll&q^{\gamma-3}\sum\limits_{u\mid q}u^{\frac{1}{2}} d(q)q^{\frac{1}{2}}\frac{q}{u}\log q \sum\limits_{r |\frac{q}{u}}\sum\limits_{d \leqslant \frac{q^{\frac{1}{4} }}{r^{\frac{1}{2}}} }( \frac{q }{d^2}+ \frac{q }{u} \log q) \\ \ll&q^{\gamma-3}\sum\limits_{u\mid q}u^{\frac{1}{2}} d(q)q^{\frac{1}{2}}\frac{q}{u}\log q \left(qd(q)+\frac{q^{\frac{5}{4} }}{u}d(q)\log q\right)\\ \ll& q^{\gamma- \frac{1}{4} }d^{3}(q) \log^3 q. \end{align}$

By the same method of $R_{211}$ ,

$\begin{align} R_{212}\ll q^{\gamma- \frac{1}{4}}d^{3}(q) \log^3 q. \end{align}$

Following from Eqs (3.10) and (3.11), estimations of $R_{211}$ and $R_{212}$ ,

$\begin{align} R_{21}\ll q^{\gamma- \frac{1}{4}}+ RP_{211}+RP_{212}\ll q^{ \gamma- \frac{1}{4}}d^{3}(q) \log^3 q. \end{align}$

(3.16)

By the similar method of $R_{12}$ and $R_{21}$ ,

$\begin{align} R_{22} = R_{221}+O(R_{222}), \end{align}$

(3.17)

where

$\begin{align} R_{221}: = &\mathop{{\sum}^{\prime}}_{n = 1}^{q} (-1)^{n+\bar{n}} \mu^{2}(n)\left( \sum\limits_{0 < |h|\leqslant H}a_{h} \left(\mathbf{e}\left(-(n+1)^{\gamma}h\right)-\mathbf{e}(-n^{\gamma}h)\right)\right) \end{align}$

and

$\begin{align} R_{222}: = &\mathop{{\sum}^{\prime}}_{n = 1}^{q} (-1)^{n+\bar{n}} \mu^{2}(n)\left( \sum\limits_{ |h|\leqslant H}b_{h} \left(\mathbf{e}\left(-(n+1)^{\gamma}h\right)+\mathbf{e}(-n^{\gamma}h)\right)\right). \end{align}$

It is obvious that

$\begin{align} R_{221} = &\mathop{{\sum}^{\prime}}_{n = 1}^{q} (-1)^{n+\bar{n}}\left(\sum\limits_{d^{2} \mid n}\mu(d)\right)\left( \sum\limits_{0 < |h|\leqslant H}a_{h} \left(\mathbf{e}\left(-(n+1)^{\gamma}h\right)-\mathbf{e}(-n^{\gamma}h)\right)\right) \\ = &\mathop{{\sum}^{\prime}}_{n = 1}^{q} (-1)^{n+\bar{n}} \mathop{\sum\limits_{d^{2} \mid n}}_{d \leqslant q^{\frac{1}{6}}}\mu(d) \left( \sum\limits_{0 < |h|\leqslant H}a_{h} \left(\mathbf{e}\left(-(n+1)^{\gamma}h\right)-\mathbf{e}(-n^{\gamma}h)\right)\right) \\ & +\mathop{{\sum}^{\prime}}_{n = 1}^{q} (-1)^{n+\bar{n}} \mathop{\sum\limits_{d^{2} \mid n}}_{q^{\frac{1}{6}} < d \leqslant q^{\frac{1}{2}} }\mu(d) \left( \sum\limits_{0 < |h|\leqslant H}a_{h} \left(\mathbf{e}\left(-(n+1)^{\gamma}h\right)-\mathbf{e}(-n^{\gamma}h)\right)\right). \end{align}$

(3.18)

From the estimate

$\mathbf{e}\left(n^{\gamma}h-(n+1)^{\gamma}h\right)-1 \ll ( n^{\gamma} -(n+1)^{\gamma})h \ll \gamma n^{\gamma -1} h,$

by partial summation,

$\begin{align} &\mathop{{\sum}^{\prime}}_{n = 1}^{q} (-1)^{n+\bar{n}} \mathop{\sum\limits_{d^{2} \mid n}}_{q^{\frac{1}{6}} < d \leqslant q^{\frac{1}{2}} }\mu(d) \left( \sum\limits_{0 < |h|\leqslant H}a_{h} \left(\mathbf{e}\left(-(n+1)^{\gamma}h\right)-\mathbf{e}(-n^{\gamma}h)\right)\right) \\ &\ll\mathop{{\sum}^{\prime}}_{n = 1}^{q} \mathop{\sum\limits_{d^{2} \mid n}}_{q^{\frac{1}{6}} < d \leqslant q^{\frac{1}{2}} } \left| \sum\limits_{0 < |h|\leqslant H}a_{h} \mathbf{e}(-n^{\gamma}h) \left(\mathbf{e}\left(n^{\gamma}h-(n+1)^{\gamma}h\right)-1\right)\right| \\ &\ll\mathop{{\sum}^{\prime}}_{n = 1}^{q} \mathop{\sum\limits_{d^{2} \mid n}}_{q^{\frac{1}{6}} < d \leqslant q^{\frac{1}{2}} } \gamma n^{\gamma-1}H \log H \\ & \ll q^{\gamma-\frac{1}{6}}H \log H . \end{align}$

(3.19)

For another term of $R_{221}$ ,

$\begin{align} &\mathop{{\sum}^{\prime}}_{n = 1}^{q} (-1)^{n+\bar{n}} \mathop{\sum\limits_{d^{2} \mid n}}_{d \leqslant q^{\frac{1}{6}} }\mu(d) \left( \sum\limits_{0 < |h|\leqslant H}a_{h} \left(\mathbf{e}\left(-(n+1)^{\gamma}h\right)-\mathbf{e}(-n^{\gamma}h)\right)\right)\\ & = \mathop{{\sum}^{\prime}}_{n = 1}^{q} (-1)^{n+\bar{n}} \mathop{\sum\limits_{d^{2} \mid n}}_{d \leqslant q^{\frac{1}{6}} }\mu(d) \left( \sum\limits_{0 < |h|\leqslant H}a_{h} \left(\mathbf{e}\left(-(n+1)^{\gamma}h\right)-\mathbf{e}(-n^{\gamma}h)\right)\right) \\ &\quad\times \left(\frac{1}{q}\sum\limits_{a = 1}^{q}\sum\limits_{s = 1}^{q}\mathbf{e}(\frac{s(m-a)}{q}) \right)\left(\frac{1}{q}\sum\limits_{b = 1}^{q}\sum\limits_{t = 1}^{q}\mathbf{e}(\frac{t(n-b)}{q})\right) \\ & = \frac{1}{q^{2}}\sum\limits_{s = 1}^{q}\sum\limits_{t = 1}^{q} \left( \mathop{\sum}_{nm \equiv 1 (\bmod q)} \mathbf{e}(\frac{sm+tn}{q}) \right)\left(\sum\limits_{a = 1}^{q-1}(-1)^{a}\mathbf{e}(-\frac{sa}{q})\right)\\ &\quad\times \left(\sum\limits_{b = 1}^{q-1}(-1)^{b}\mathbf{e}(-\frac{tb}{q}) \mathop{\sum\limits_{d^{2} \mid b}}_{d \leqslant q^{\frac{1}{6} } }\mu(d) \sum\limits_{0 < |h|\leqslant H}a_{h} \left(\mathbf{e}\left(-(b+1)^{\gamma}h\right)-\mathbf{e}(-b^{\gamma}h)\right) \right ). \end{align}$

(3.20)

We just need to give an estimation of the last part in (3.20). Similarly, let

$g (x) = \mathbf{e}\left(\left((d^{2}x)^{\gamma}-(d^{2}x+1)^{\gamma}\right)h \right)-1,$

then

$\begin{align} g (x) &\ll\ |h| (d^2x)^{\gamma-1}, \\ \frac{\partial g(x)}{\partial x} &\ll |h| d^{2\gamma-2}x^{\gamma-2}. \end{align}$

By partial summation,

$\begin{align} &\sum\limits_{b = 1}^{q-1}(-1)^{b}\mathbf{e}(-\frac{tb}{q}) \mathop{\sum\limits_{d^{2} \mid b}}_{d \leqslant q^{\frac{1}{6} } }\mu(d) \sum\limits_{0 < |h|\leqslant H}a_{h} \left(\mathbf{e}\left(-(b+1)^{\gamma}h\right)-\mathbf{e}(-b^{\gamma}h)\right) \\ & = \sum\limits_{0 < |h|\leqslant H}a_{h} \sum\limits_{d \leqslant q^{ \frac{1}{6} }}\mu(d)\sum\limits_{1 \leqslant b \leqslant \lfloor\frac{q-1}{d^2}\rfloor} \mathbf{e}\left((\frac{d^{2}}{2}-\frac{td^{2}}{q})b-(d^{2}b)^{\gamma}h\right)g(b), \\ & \ll\sum\limits_{0 < |h|\leqslant H}|h|^{-1} \sum\limits_{d \leqslant q^{ \frac{1}{6} }} \left|\int_{ 1}^{ \lfloor\frac{q-1}{d^2}\rfloor} g(x) \mathrm{d}\left(\sum\limits_{1 < b \leqslant x} \mathbf{e}\left((\frac{d^{2}}{2}-\frac{td^{2}}{q})b-(d^{2}b)^{\gamma}h\right) \right) \right|\\ & \ll\sum\limits_{0 < |h|\leqslant H}|h|^{-1} \sum\limits_{d \leqslant q^{ \frac{1}{6} }} \left|g( \lfloor\frac{q-1}{d^2}\rfloor) \sum\limits_{ 1 \leqslant b \leqslant \lfloor\frac{q-1}{d^2}\rfloor} \mathbf{e}\left((\frac{d^{2}}{2}-\frac{td^{2}}{q})b-(d^{2}b)^{\gamma}h\right)\right|\\ &\quad +\sum\limits_{0 < |h|\leqslant H}|h|^{-1} \sum\limits_{d \leqslant q^{ \frac{1}{6} }} \left|\int_{ 1}^{ \lfloor\frac{q-1}{d^2}\rfloor}\frac{\partial g(x)}{\partial x} \sum\limits_{ 1 \leqslant b \leqslant x} \mathbf{e}\left((\frac{d^{2}}{2}-\frac{td^{2}}{q})b-(d^{2}b)^{\gamma}h\right) \mathrm{d}x\right|, \end{align}$

where

$g( \lfloor\frac{q-1}{d^2}\rfloor) \ll |h|q^{\gamma-1}$

and

$\begin{align} &\sum\limits_{ 1 \leqslant b \leqslant \lfloor\frac{q-1}{d^2}\rfloor} \mathbf{e}\left((\frac{d^{2}}{2}-\frac{td^{2}}{q})b-(d^{2}b)^{\gamma}h\right) \\ & = \sum\limits_{ 1 \leqslant b \leqslant q^{\frac{1}{6}} } \mathbf{e}\left((\frac{d^{2}}{2}-\frac{td^{2}}{q})b-(d^{2}b)^{\gamma}h\right)+\sum\limits_{ q^{\frac{1}{6}} < b \leqslant \lfloor\frac{q-1}{d^2}\rfloor} \mathbf{e}\left((\frac{d^{2}}{2}-\frac{td^{2}}{q})b-(d^{2}b)^{\gamma}h\right). \end{align}$

It is obvious that

$\begin{align} \sum\limits_{ 1 \leqslant b \leqslant q^{\frac{1}{6}} } \mathbf{e}\left((\frac{d^{2}}{2}-\frac{td^{2}}{q})b-(d^{2}b)^{\gamma}h\right)\ll q^{\frac{1}{6}}. \end{align}$

Suppose $q$ be large enough and for $b > q^{\frac{1}{6}}$ , when $2 \nmid d$ or $q \nmid td^2$ ,

$\left\| (\frac{d^{2}}{2}-\frac{td^{2}}{q})-\gamma d^{2\gamma}b^{\gamma-1}h\right\|^{-1} \geqslant \frac{1}{2} \left\| (\frac{1}{2}-\frac{t }{q})d^{2}\right\|^{-1} > 0,$

and applying Lemma 2.6, we have

$\begin{align} \sum\limits_{q^{\frac{1}{6}} < b \leqslant \lfloor\frac{q-1}{d^2}\rfloor} \mathbf{e}\left((\frac{d^{2}}{2}-\frac{td^{2}}{q})b-(d^{2}b)^{\gamma}h\right) \ll& \left\| (\frac{d^{2}}{2}-\frac{td^{2}}{q})-\gamma d^{2\gamma}b^{\gamma-1}h\right\|^{-1} \\ \ll& \left\| (\frac{1}{2}-\frac{t }{q})d^{2}\right\|^{-1} . \end{align}$

$\sum\limits_{ 1 \leqslant b \leqslant \lfloor\frac{q-1}{d^2}\rfloor} \mathbf{e}\left((\frac{d^{2}}{2}-\frac{td^{2}}{q})b-(d^{2}b)^{\gamma}h\right) \ll \begin{cases} q^{\frac{1}{6}}+ \left\| (\frac{1}{2}-\frac{t }{q})d^{2}\right\|^{-1} , & 2 \nmid d \text{ or } q \nmid td^2;\\ \frac{q}{d^2}, & 2 \mid d \text{ and } q \mid td^2 ;\end{cases}$

which means

We denote

$\begin{align} T(c)&: = \sum\limits_{d\leqslant q^{\frac{1}{6} }} \# \left\{ (\frac{q}{u}-2t)d^{2} \equiv c (\bmod 2\frac{q}{u}), t\leqslant \frac{q}{u}\right\} \\ &\ll \sum\limits_{d\leqslant q^{\frac{1}{6} }} (\frac{q}{u}, d^{2}) \\ &\ll q^{\frac{1}{3} }d(q), \end{align}$

thus,

$\begin{align} &\mathop{{\sum}^{\prime}}_{n = 1}^{q} (-1)^{n+\bar{n}} \mathop{\sum\limits_{d^{2} \mid n}}_{d \leqslant q^{\frac{1}{6}} }\mu(d) \left( \sum\limits_{0 < |h|\leqslant H}a_{h} \left(\mathbf{e}\left(-(n+1)^{\gamma}h\right)-\mathbf{e}(-n^{\gamma}h)\right)\right)\\ & \ll q^{-2}\sum\limits_{s = 1}^{q}\sum\limits_{t = 1}^{q}(s,t,q)^{\frac{1}{2}}d(q)q^{\frac{1}{2}}\frac{H q^{ \gamma- 1}}{|\cos\frac{s}{q}\pi|}\left(q^{\frac{1}{ 3}}+ \mathop{\sum\limits_{d \leqslant q^{ \frac{1}{6} }}}_{2 \nmid d \text{ or } q \nmid td^2} \left\| (\frac{1}{2}-\frac{t }{q})d^{2}\right\|^{-1}\right) \\ &\quad+q^{-2}\sum\limits_{s = 1}^{q}\sum\limits_{t = 1}^{q}(s,t,q)^{\frac{1}{2}}d(q)q^{\frac{1}{2}}\frac{H q^{ \gamma}}{|\cos\frac{s}{q}\pi|}\mathop{\sum\limits_{d \leqslant q^{ \frac{1}{6} }}}_{2 \mid d \text{ and } q \mid td^2}d^{-2} \\ & \ll Hq^{ \gamma-\frac{5}{2}}\sum\limits_{u\mid q}u^{\frac{1}{2}} d(q) \sum\limits_{s = 1}^{\frac{q}{u}}\frac{1}{|1-2 \frac{su}{q} |} \sum\limits_{t = 1}^{\frac{q}{u}} \left(q^{\frac{1}{3}}+ \sum\limits_{d \leqslant q^{\frac{1}{6} }} \left\| (\frac{1}{2}-\frac{ut }{q})d^{2}\right\|^{-1} \right) \\ &\quad+Hq^{ \gamma-\frac{3}{2}} \sum\limits_{u\mid q}u^{\frac{1}{2}} d(q) \sum\limits_{s = 1}^{\frac{q}{u}}\frac{1}{|1-2 \frac{su}{q} |} \sum\limits_{d \leqslant q^{ \frac{1}{6} }}\mathop{\sum\limits_{t = 1}^{\frac{q}{u}}}_{ q \mid td^2}d^{-2} \\ & \ll H q^{ \gamma-\frac{1}{6}} d^2(q) \log q+ Hq^{ \gamma-\frac{1}{3}}d^2(q) \log q \\ &\quad+ Hq^{ \gamma- \frac{3}{2}}\sum\limits_{u\mid q}u^{-\frac{1}{2}} d(q) \log^{2}q \max\limits_{C} \sum\limits_{C < c\leqslant2C} \|\frac{C}{2\frac{q}{u}}\|^{-1} T(c) \\ &\ll Hq^{ \gamma -\frac{1}{6}}d^2(q) \log^2 q. \end{align}$

(3.21)

With Eqs (3.18) and (3.19), we have

$\begin{align} R_{221}&\ll Hq^{ \gamma -\frac{1}{6}}d^2(q) \log^2 q+H q^{ \gamma -\frac{1}{6}} \log H. \end{align}$

(3.22)

For $R_{222}$ , the contribution from $h = 0$ can be bounded by similar methods of $R_{21}$ , and the contribution from $h \neq 0$ can be bounded by similar methods of $R_{221}$ . Taking

$H = \log q ,$

we obtain

$\begin{align} R_{222}& = b_{0} \mathop{{\sum}^{\prime}}_{n = 1}^{q} (-1)^{n+\bar{n}} \mu^{2}(n) +\mathop{{\sum}^{\prime}}_{n = 1}^{q} (-1)^{n+\bar{n}}\mu^{2}(n)\left(\sum\limits_{0 < |h|\leqslant H}b_{h} \left(e\left(-(n+1)^{\gamma}h\right)+e\left(-n^{\gamma}h\right)\right)\right) \\ &\ll H^{-1}q^{\frac{3}{4}}d^{3}(q)\log^2 q +Hq^{ \gamma -\frac{1}{6}}d^2(q) \log^2 q\\ &\ll q^{\frac{3}{4}}d^{3}(q)\log q+ q^{ \gamma -\frac{1}{6}}d^{2}(q) \log^3 q . \end{align}$

(3.23)

Following from Eqs (3.16), (3.22), and (3.23),

$\begin{align} R_{2}\ll q^{\gamma-\frac{1}{6}} d^{2}(q) \log^3 q+ q^{\frac{3}{4}}d^{3}(q)\log q . \end{align}$

(3.24)

Hence, from Eqs (3.1), (3.8), and (3.24), we derive that

$\begin{align} R(c;q) = &\frac{3}{\pi^{2}}\prod\limits_{p\mid q}(1+p^{-1})^{-1}q^{\gamma}+O\left( \sum\limits_{p\mid q}(1-p^{-\frac{1}{2}}) ^{-1} q^{\gamma-\frac{1}{2}}\right)\\ &+O\left(q^{\frac{7}{13}\gamma+\frac{4}{13}}\prod\limits_{p\mid q}(1-p^{-\frac{1}{2}})^{-1}\log q\right)+O\left( q^{\frac{3}{4}}d^{3}(q)\log q \right)+O(q^{ \gamma -\frac{1}{6}} d^{2}(q)\log^3 q). \end{align}$

We need the error terms to be smaller than the main term, so

$\begin{cases} \frac{7}{13}\gamma+\frac{4}{13} < \gamma ,\\ \frac{3}{4} < \gamma ,\end{cases}$

which means the range of $c$ is $(1, \frac{4}{3})$ . The reason why the range of $c$ is changed is that $R(c; q)$ requires $q$ large enough.

4. Conclusions

In this paper, we generalize the Lehmer problem by considering the count of square-free numbers in the intersection of the Lehmer set and Piatetski-Shapiro sequence when $q$ is an odd integer and large enough. By methods of exponential sum and Kloosterman sum, we study its asymptotic properties and give a sharp asymptotic formula as $q$ tends to infinity.

Based on this result, we will consider some distribution problems similar to the Lehmer problem with more special sequences, which is significant for understanding the distribution properties of those problems.

Author contributions

Xiaoqing Zhao: calculations, writing and editing; Yuan Yi: methodology and reviewing. All authors have read and agreed to the published version of the manuscript.

Use of AI tools declaration

In preparing this manuscript, we employed the language model ChatGPT-4 for the purpose of grammatical corrections. It did not influence the calculations and conclusion in this paper.

Acknowledgments

This work is supported by Natural Science Foundation No.12271422 of China. The authors would like to express their gratitude to the referee for very helpful and detailed comments.

Conflict of interest

The authors declare that there are no conflicts of interest regarding the publication of this paper.

References

[1]	T. Wang, P. Xuan, Z. Liu, T. Zhang, Assistant diagnosis with Chinese electronic medical records based on CNN and BiLSTM with phrase-level and word-level attentions, BMC Bioinf. , 21 (2020). https://doi.org/10.1186/s12859-020-03554-x doi: 10.1186/s12859-020-03554-x
[2]	J. Tsai, G. Bond, A comparison of electronic records to paper records in mental health centers, Int. J. Qual. Health Care, 20 (2008), 136–143. https://doi.org/10.1093/intqhc/mzm064 doi: 10.1093/intqhc/mzm064
[3]	Y. Hu, Research on the information diagnostic technology based on medical information, University of Electronic Science and Technology of China, 2015.
[4]	Z. Obermeyer, E. J. Emanuel, Predicting the future—big data, machine learning, and clinical medicine, N. Engl. J. Med. , 375 (2016), 1216–1219. https://doi.org/10.1056/NEJMp1606181 doi: 10.1056/NEJMp1606181
[5]	Y. LeCun, Y. Bengio, G. Hinton, Deep learning, Nature, 521 (2015), 436–444. https://doi.org/10.1038/nature14539 doi: 10.1038/nature14539
[6]	J. Yang, Y. Guan, B. He, C. Qu, Q. Yu, Y. Liu, et al., Corpus construction for named entities and entity relations on chinese electronic medical records, J. Softw. , 27 (2016), 2725–2746.
[7]	L. R. Rabiner, A tutorial on hidden Markov models and selected applications in speech recognition, Proc. IEEE, 77 (1989), 257–286. https://doi.org/10.1109/5.18626 doi: 10.1109/5.18626
[8]	A. Roberts, R. Gaizauskas, M. Hepple, Extracting clinical relationships from patient narratives, in Proceedings of the Workshop on Current Trends in Biomedical Natural Language Processing, (2008), 10–18. https://doi.org/10.3115/1572306.1572309
[9]	J. Lafferty, A. McCallum, F. C. N. Pereira, Conditional random fields: Probabilistic models for segmenting and labeling sequence data, in Proceedings of the 18th International Conference on Machine Learning 2001 (ICML 2001), (2001), 282–289. https://repository.upenn.edu/handle/20.500.14332/6188
[10]	S. Hochreiter, J. Schmidhuber, Long short-term memory, Neural Comput. , 9 (1997), 1735–1780.
[11]	J. Devlin, M. W. Chang, K. Lee, K. Toutanova, BERT: Pre-training of deep bidirectional transformers for language understanding, preprint, arXiv: 1810.048052018.
[12]	M. E. Peters, M. Neumann, M. Iyyer, M. Gardner, C. Clark, K. Lee, et al., Deep contextualized word representations, Assoc. Comput. Linguist. , 1 (2018), 2227–2237. https://doi.org/10.18653/v1/N18-1202 doi: 10.18653/v1/N18-1202
[13]	T. Younga, D. Hazarikab, S. Poriac, E. Cambriad, Recent trends in deep learning based natural language processing, IEEE Comput. Intell. Mag. , 13 (2018), 55–75. https://doi.org/10.1109/MCI.2018.2840738 doi: 10.1109/MCI.2018.2840738
[14]	L. Ouyang, Y. Tian, H. Tang, B. Zhang, Chinese named entity recognition based on B-LSTM neural network with additional features, in International Conference on Security, Privacy and Anonymity in Computation, Communication and Storage, (2017), 269–279. https://doi.org/10.1007/978-3-319-72389-1_22
[15]	Y. Xiang, Chinese named entity recognition with character-word mixed embedding, in Proceedings of the 2017 ACM on Conference on Information and Knowledge Management, (2017), 2055–2058.
[16]	H. Yang, H. Gao, Toward sustainable virtualized healthcare: Extracting medical entities from Chinese online health consultations using deep neural networks, Sustainability, 10 (2018), 3292. https://doi.org/10.3390/su10093292 doi: 10.3390/su10093292
[17]	W. Zhang, S. Jiang, S. Zhao, K. Hou, Y. Liu, L. Zhang, A BERT-BiLSTM-CRF model for Chinese electronic medical records named entity recognition, in 2019 12th International Conference on Intelligent Computation Technology and Automation (ICICTA), (2019), 166–169. https://doi.org/10.1109/ICICTA49267.2019.00043
[18]	X. Zhang, Y. Zhang, Q. Zhang, Y. Ren, T. Qiu, J. Ma, et al., Extracting comprehensive clinical information for breast cancer using deep learning methods, Int. J. Med. Inf. , 132 (2019), 103985.
[19]	L. Li, L. Jin, Y. Jiang, D. Huang, Recognizing biomedical named entities based on the sentence vector/twin word embeddings conditioned bidirectional LSTM, in Chinese Computational Linguistics and Natural Language Processing Based on Naturally Annotated Big Data, (2016), 165–176. https://doi.org/10.1007/978-3-319-47674-2_15
[20]	M. Habibi, L. Weber, M. Neves, D. L. Wiegandt, U. Leser, Deep learning with word embeddings improves biomedical named entity recognition, Bioinformatics, 33 (2017), i37–i48. https://doi.org/10.1093/bioinformatics/btx228 doi: 10.1093/bioinformatics/btx228
[21]	J. P. C. Chiu, E. Nichols, Named entity recognition with bidirectional LSTM-CNNs, Trans. Assoc. Comput. Linguist., 4 (2016), 357–370. https://doi.org/10.1162/tacl_a_00104 doi: 10.1162/tacl_a_00104
[22]	L. Li, Y. Guo, Biomedical named entity recognition with CNN-BLSTM-CRF, J. Chin. Inf. Newsp., (2018), 116–122.
[23]	D. S. Sachan, P. Xie, M. Sachan, P. Xing, Effective use of bidirectional language modeling for transfer learning in biomedical named entity recognition, in Proceedings of the 3rd Machine Learning for Healthcare Conference, (2018), 383–402.
[24]	E. F. Tjong K. Sang, J. Veenstra, in Proceedings of the Ninth Conference on European Chapter of the Association for Computational Linguistics, (1999), 173–179. https://doi.org/10.3115/977035.977059
[25]	X. Dong, S. Chowdhury, L. Qian, Y. Guan, J. Yang, Q. Yu, Transfer bi-directional LSTM rnn for named entity recognition in chinese electronic medical records, in 2017 IEEE 19th International Conference on e-Health Networking, Applications and Services (Healthcom), (2017), 12–15. https://doi.org/10.1109/HealthCom.2017.8210840

Reader Comments

Your name:*

Email:*
© 2024 the Author(s), licensee AIMS Press. This is an open access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/4.0)