Latex

This is my first time to use latex to edit a paper.  I have to say the process is almost acceptable. 

\documentclass[times, twoside]{zHenriquesLab-StyleBioRxiv}
%\usepackage{blindtext}
\usepackage{ulem}
\usepackage{url}
% Please give the surname of the lead author for the running footer
%\leadauthor{Henriques} 
\begin{document}

\title{Research Proposal}
\shorttitle{My Research Proposal}

% Use letters for affiliations, numbers to show equal authorship (if applicable) and to indicate the corresponding author
\author{Changqi Sun}

\affil[]{A novel supervised algorithm for sleep/wake identification using simulated actigraphy activity counts}

\maketitle


%TC:break Abstract
%the command above serves to have a word count for the abstract
\begin{abstract}
This research proposal is based on "\emph{A novel machine learning unsupervised algorithm for sleep/wake identification using actigraphy}" to address a problem improving the discrimination accuracy of sleep/wake state. 
\end {abstract}
%TC:break main
%the command above serves to have a word count for the abstract

%\begin{keywords}
%bla | bla | bla | bla
%\end{keywords}

\begin{corrauthor}
\texttt{sunchangqi\_sdu{@}163.com}
%$sunchangqi$_$sdu\at 163.com$
\end{corrauthor}

\section*{Introduction}
From my perspective, the identification of sleep and wake using actigraphy plays an essential role in the promotion of wearable devices into the medical detection field. In addition, I know the miseries from the insomnia, because I have once suffered from that. There are many methods to solve this discrimination problem such as logistic regression representing supervised algorithm and Hidden Markov Model(HMM) as  representation of unsupervised algorithm(Xinyue et al.2020\cite{2020A}). This paper also compared the unsupervised algorithm embedded in the Actiwatch Software(AS) with pre-trained supervised UCSD algorithm. In conclusion, the accuracy of HMM, AS and UCSD is 85.7\%, 84.7\% and 85\% respectively. It is easy to find that the performance of three algorithms is neck to neck. The main reason behind the situation , in my opinion, is that they do not take the fluctuations into consideration. Therefore, if we would like to further improve the discrimination accuracy, we need an algorithm that takes the deviation into account.

\section*{Methods}
In this research proposal, I proposed a more accurate sleep and wake identification method based on moving average and standard deviation model. The moving averages usually are used to illustrate the increasing or decreasing trend of stocks. This model could be described as follows:
$$
\left\{s_{1}, s_{2}\right\}=f\left(\mu_{1}, \mu_{2}, \sigma\right)
$$
where $s_{1}$ and $s_{2}$ denote the sleep state and the wake state respectively, with $\mu_{1}$ denoting moving average of activity counts in short window, $\mu_{2}$ denoting moving average of activity counts in long window and $\sigma$ denoting standard deviation of activity counts in short window. According to the information in collection of activity counts, the value of standard deviation in sleep or wake state is almost stable, but in the transformation period will take a leap. So the points with high standard deviation value are the boundaries between sleep and wake state. Therefore, we could use these points as the boundaries to discriminate two states. Furthermore, there is coming another problem how to identify which period is sleep or wake state, which could be addressed by the moving averages in short and long window. Behind the boundary, if the moving average in short window crosses up through in long window, the period followed by the boundary is wake state. In contrast, the period followed by the boundary will be sleep state with moving average in short window crossing down through in long window.

\section*{Experimental}
To test whether my model works, I did a little experiment containing three steps:

Step 1: Generate a series of activity counts with 30 values in random way: the first 10 values between 10.1 and 14.4; the middle 9 values between 0 and 1; the last 11 values between 10.1 and 14.4.

Step 2: Calculate the moving averages with window = 3 and window =10 and moving standard deviation with window = 3 (The final result of moving standard deviation is divided by 10 to make the graph more intuitive).

Step 3: Depict the two moving averages and moving standard deviation line.

The implementation of this model is in Python. (Its codes link:

\url{https://blog.csdn.net/weixin_38396940/article/details/117553693}
\begin{figure}[htb]
\centering
\includegraphics[width=.8\linewidth]{Figures/Figure_1}
\caption{The red line denotes the moving average with window $=$ 3; the blue line denotes the moving average with window $=$ 10; the black line denotes the moving standard deviation divided by 10 with window =3.}
%\label{fig:computerNo}
\end{figure}
\subsection*{Results} 
As you can see, the black line has two bulges that are in line with the transformation of the two states. In addition, the red line crosses down through the blue line representing the next period is sleep state. The parameters discussed further of this model refer to value of threshold describing the sleep or wake period and windows of moving averages and standard deviation.     

\section*{Bibliography}
\bibliography{zHenriquesLab-Mendeley}

\end{document}

标签：deviation,Latex,state,window,wake,moving,sleep
来源： https://blog.csdn.net/weixin_38396940/article/details/117607964