284 lines
8.7 KiB
TeX
284 lines
8.7 KiB
TeX
% NeurIPS Conference Paper Template
|
|
% For submission to Neural Information Processing Systems (NeurIPS)
|
|
% Last updated: 2024
|
|
% Note: Use the official neurips_2024.sty file from the conference website
|
|
|
|
\documentclass{article}
|
|
|
|
% Required packages (neurips_2024.sty provides most formatting)
|
|
\usepackage{neurips_2024} % Official NeurIPS style file (download from conference site)
|
|
|
|
% Recommended packages
|
|
\usepackage{amsmath}
|
|
\usepackage{amssymb}
|
|
\usepackage{amsthm}
|
|
\usepackage{graphicx}
|
|
\usepackage{algorithm}
|
|
\usepackage{algorithmic}
|
|
\usepackage{hyperref}
|
|
\usepackage{url}
|
|
\usepackage{booktabs} % For better tables
|
|
\usepackage{multirow}
|
|
\usepackage{microtype} % Improved typography
|
|
|
|
% Theorems, lemmas, etc.
|
|
\newtheorem{theorem}{Theorem}
|
|
\newtheorem{lemma}{Lemma}
|
|
\newtheorem{proposition}{Proposition}
|
|
\newtheorem{corollary}{Corollary}
|
|
\newtheorem{definition}{Definition}
|
|
|
|
% Title and Authors
|
|
\title{Your Paper Title: Concise and Descriptive \\ (Maximum Two Lines)}
|
|
|
|
% Authors - ANONYMIZED for initial submission
|
|
% For initial submission (double-blind review):
|
|
\author{
|
|
Anonymous Authors \\
|
|
Anonymous Institution(s) \\
|
|
}
|
|
|
|
% For camera-ready version (after acceptance):
|
|
% \author{
|
|
% First Author \\
|
|
% Department of Computer Science \\
|
|
% University Name \\
|
|
% City, State, Postal Code \\
|
|
% \texttt{first.author@university.edu} \\
|
|
% \And
|
|
% Second Author \\
|
|
% Company/Institution Name \\
|
|
% Address \\
|
|
% \texttt{second.author@company.com} \\
|
|
% \And
|
|
% Third Author \\
|
|
% Institution \\
|
|
% \texttt{third.author@institution.edu}
|
|
% }
|
|
|
|
\begin{document}
|
|
|
|
\maketitle
|
|
|
|
\begin{abstract}
|
|
Write a concise abstract (150-250 words) summarizing your contributions. The abstract should clearly state: (1) the problem you address, (2) your approach/method, (3) key results/findings, and (4) significance/implications. Make it accessible to a broad machine learning audience.
|
|
\end{abstract}
|
|
|
|
\section{Introduction}
|
|
\label{sec:introduction}
|
|
|
|
Introduce the problem you're addressing and its significance in machine learning or AI. Motivate why this problem is important and challenging.
|
|
|
|
\subsection{Background and Motivation}
|
|
Provide necessary background for understanding your work. Explain the gap in current methods or knowledge.
|
|
|
|
\subsection{Contributions}
|
|
Clearly state your main contributions as a bulleted list:
|
|
\begin{itemize}
|
|
\item First contribution: e.g., "We propose a novel architecture for..."
|
|
\item Second contribution: e.g., "We provide theoretical analysis showing..."
|
|
\item Third contribution: e.g., "We demonstrate state-of-the-art performance on..."
|
|
\end{itemize}
|
|
|
|
\subsection{Paper Organization}
|
|
Briefly describe the structure of the remainder of the paper.
|
|
|
|
\section{Related Work}
|
|
\label{sec:related}
|
|
|
|
Discuss relevant prior work and how your work differs. Organize by themes or approaches rather than chronologically. Be fair and accurate in describing others' work.
|
|
|
|
Cite key papers \cite{lecun2015deep, vaswani2017attention, devlin2019bert} and explain how your work builds upon or differs from them.
|
|
|
|
\section{Problem Formulation}
|
|
\label{sec:problem}
|
|
|
|
Formally define the problem you're solving. Include mathematical notation and definitions.
|
|
|
|
\subsection{Notation}
|
|
Define your notation clearly. For example:
|
|
\begin{itemize}
|
|
\item $\mathcal{X}$: input space
|
|
\item $\mathcal{Y}$: output space
|
|
\item $f: \mathcal{X} \rightarrow \mathcal{Y}$: function to learn
|
|
\item $\mathcal{D} = \{(x_i, y_i)\}_{i=1}^n$: training dataset
|
|
\end{itemize}
|
|
|
|
\subsection{Objective}
|
|
State your learning objective formally, e.g.:
|
|
\begin{equation}
|
|
\min_{\theta} \mathbb{E}_{(x,y) \sim \mathcal{D}} \left[ \mathcal{L}(f_\theta(x), y) \right]
|
|
\end{equation}
|
|
where $\mathcal{L}$ is the loss function and $\theta$ are model parameters.
|
|
|
|
\section{Method}
|
|
\label{sec:method}
|
|
|
|
Describe your proposed method in detail. This is the core technical contribution of your paper.
|
|
|
|
\subsection{Model Architecture}
|
|
Describe the architecture of your model with sufficient detail for reproduction. Include figures if helpful.
|
|
|
|
\begin{figure}[t]
|
|
\centering
|
|
% \includegraphics[width=0.8\textwidth]{architecture.pdf}
|
|
\caption{Model architecture diagram. Describe the key components and data flow. Use colorblind-safe colors.}
|
|
\label{fig:architecture}
|
|
\end{figure}
|
|
|
|
\subsection{Training Procedure}
|
|
Explain how you train the model, including:
|
|
\begin{algorithm}[t]
|
|
\caption{Training Algorithm}
|
|
\label{alg:training}
|
|
\begin{algorithmic}[1]
|
|
\STATE \textbf{Input:} Training data $\mathcal{D}$, learning rate $\alpha$
|
|
\STATE \textbf{Output:} Trained parameters $\theta$
|
|
\STATE Initialize $\theta$ randomly
|
|
\FOR{epoch $= 1$ to $T$}
|
|
\FOR{batch $(x, y)$ in $\mathcal{D}$}
|
|
\STATE Compute loss: $\mathcal{L} = \mathcal{L}(f_\theta(x), y)$
|
|
\STATE Update: $\theta \leftarrow \theta - \alpha \nabla_\theta \mathcal{L}$
|
|
\ENDFOR
|
|
\ENDFOR
|
|
\RETURN $\theta$
|
|
\end{algorithmic}
|
|
\end{algorithm}
|
|
|
|
\subsection{Key Components}
|
|
Describe key technical innovations or components in detail.
|
|
|
|
\section{Theoretical Analysis}
|
|
\label{sec:theory}
|
|
|
|
If applicable, provide theoretical analysis of your method.
|
|
|
|
\begin{theorem}
|
|
\label{thm:main}
|
|
State your main theoretical result here.
|
|
\end{theorem}
|
|
|
|
\begin{proof}
|
|
Provide proof or sketch of proof. Full proofs can go in the appendix.
|
|
\end{proof}
|
|
|
|
\section{Experiments}
|
|
\label{sec:experiments}
|
|
|
|
Describe your experimental setup and results.
|
|
|
|
\subsection{Experimental Setup}
|
|
\textbf{Datasets:} Describe datasets used (e.g., ImageNet, CIFAR-10, etc.).
|
|
|
|
\textbf{Baselines:} List baseline methods for comparison.
|
|
|
|
\textbf{Implementation Details:} Provide implementation details including hyperparameters, hardware, training time.
|
|
|
|
\textbf{Evaluation Metrics:} Define metrics used (accuracy, F1, AUC, etc.).
|
|
|
|
\subsection{Main Results}
|
|
Present your main experimental results.
|
|
|
|
\begin{table}[t]
|
|
\centering
|
|
\caption{Performance comparison on benchmark datasets. Bold indicates best performance. Results reported as mean ± std over 3 runs.}
|
|
\label{tab:main_results}
|
|
\begin{tabular}{lcccc}
|
|
\toprule
|
|
Method & Dataset 1 & Dataset 2 & Dataset 3 & Average \\
|
|
\midrule
|
|
Baseline 1 & 85.3 ± 0.5 & 72.1 ± 0.8 & 90.2 ± 0.3 & 82.5 \\
|
|
Baseline 2 & 87.2 ± 0.4 & 74.5 ± 0.6 & 91.1 ± 0.5 & 84.3 \\
|
|
\textbf{Our Method} & \textbf{91.7 ± 0.3} & \textbf{79.8 ± 0.5} & \textbf{94.3 ± 0.2} & \textbf{88.6} \\
|
|
\bottomrule
|
|
\end{tabular}
|
|
\end{table}
|
|
|
|
\subsection{Ablation Studies}
|
|
Conduct ablation studies to understand which components contribute to performance.
|
|
|
|
\subsection{Analysis}
|
|
Provide deeper analysis of results, failure cases, limitations, etc.
|
|
|
|
\section{Discussion}
|
|
\label{sec:discussion}
|
|
|
|
Discuss your findings, limitations, and broader implications.
|
|
|
|
\subsection{Limitations}
|
|
Honestly acknowledge limitations of your work.
|
|
|
|
\subsection{Broader Impacts}
|
|
Discuss potential positive and negative societal impacts (required by NeurIPS).
|
|
|
|
\section{Conclusion}
|
|
\label{sec:conclusion}
|
|
|
|
Summarize your main contributions and findings. Suggest future research directions.
|
|
|
|
% Acknowledgments (add after acceptance, not in submission version)
|
|
\section*{Acknowledgments}
|
|
Thank collaborators, funding sources (with grant numbers), and compute resources. Not included in double-blind submission.
|
|
|
|
% References
|
|
\bibliographystyle{plainnat} % or other NeurIPS-compatible style
|
|
\bibliography{references} % Your .bib file
|
|
|
|
% Appendix (optional, unlimited pages)
|
|
\appendix
|
|
|
|
\section{Additional Proofs}
|
|
\label{app:proofs}
|
|
|
|
Provide full proofs of theorems here.
|
|
|
|
\section{Additional Experimental Results}
|
|
\label{app:experiments}
|
|
|
|
Include additional experiments, more ablations, qualitative results, etc.
|
|
|
|
\section{Hyperparameters}
|
|
\label{app:hyperparameters}
|
|
|
|
List all hyperparameters used in experiments for reproducibility.
|
|
|
|
\begin{table}[h]
|
|
\centering
|
|
\caption{Hyperparameters used in all experiments}
|
|
\begin{tabular}{ll}
|
|
\toprule
|
|
Hyperparameter & Value \\
|
|
\midrule
|
|
Learning rate & 0.001 \\
|
|
Batch size & 64 \\
|
|
Optimizer & Adam \\
|
|
Weight decay & 0.0001 \\
|
|
Epochs & 100 \\
|
|
\bottomrule
|
|
\end{tabular}
|
|
\end{table}
|
|
|
|
\section{Code and Data}
|
|
\label{app:code}
|
|
|
|
Provide links to code repository (anonymized for review, e.g., anonymous GitHub):
|
|
\begin{itemize}
|
|
\item Code: \url{https://anonymous.4open.science/r/project-XXXX}
|
|
\item Data: Available upon request / at [repository]
|
|
\end{itemize}
|
|
|
|
\end{document}
|
|
|
|
% Notes for Authors:
|
|
% 1. Main paper: 8 pages maximum (excluding references and appendix)
|
|
% 2. References: unlimited pages
|
|
% 3. Appendix: unlimited pages (reviewed at discretion of reviewers)
|
|
% 4. Use double-blind anonymization for initial submission
|
|
% 5. Include broader impact statement
|
|
% 6. Code submission strongly encouraged (anonymous for review)
|
|
% 7. Use official neurips_2024.sty file (download from NeurIPS website)
|
|
% 8. Font: Times, 10pt (enforced by style file)
|
|
% 9. Figures should be colorblind-friendly
|
|
% 10. Ensure reproducibility: report seeds, hyperparameters, dataset splits
|
|
|