检索结果-内蒙古大学图书馆

A Stein’s Method Approach to the Linear Noise Approximation for Stationary Distributions of Chemical Reaction Networks

学校读者我要写书评

暂无评论

arXiv 2023年

作者： Grunberg, Theodore W. Vecchio, Domitilla Del Department of Electrical Engineering and Computer Science MIT CambridgeMA02139 United States Department of Mechanical Engineering MIT CambridgeMA02139 United States

Stochastic Chemical Reaction Networks are continuous time Markov chain models that describe the time evolution of the molecular counts of species interacting stochastically via discrete reactions. Such models are ubiquitous in systems and synthetic biology, but often have a large or infinite number of states, and thus are not amenable to computation and analysis. Due to this, approximations that rely on the molecular counts and the volume being large are commonly used, with the most common being the Reaction Rate Equations and the Linear Noise Approximation. For finite time intervals, Kurtz established the validity of the Reaction Rate Equations and Linear Noise Approximation, by proving law of large numbers and central limit theorem results respectively. However, the analogous question for the stationary distribution of the Markov chain model has remained mostly unanswered, except for chemical reaction networks with special structures or bounded molecular counts. In this work, we use Stein’s Method to obtain sufficient conditions for the stationary distribution of an appropriately scaled Stochastic Chemical Reaction Network to converge to the Linear Noise Approximation as the system size goes to infinity. Our results give non asymptotic bounds on the error in the Reaction Rate Equations and in the Linear Noise Approximation as applied to the stationary distribution. As a special case, we give conditions under which the global exponential stability of an equilibrium point of the Reaction Rate Equations is sufficient to obtain our error bounds, thus permitting one to obtain conclusions about the Markov chain by analyzing the deterministic Reaction Rate *** Codes 60J28, 60F05 (Primary) 92C40 (Secondary) Copyright © 2023, The Authors. All rights reserved.

关键词： Reaction rates

An algorithmic framework for synthetic cost-aware decision making in molecular design

学校读者我要写书评

暂无评论

arXiv 2023年

作者： Fromer, Jenna C. Coley, Connor W. Department of Chemical Engineering MIT CambridgeMA02139 United States Department of Electrical Engineering and Computer Science MIT CambridgeMA02139 United States

Small molecules exhibiting desirable property profiles are often discovered through an iterative process of designing, synthesizing, and testing sets of molecules. The selection of molecules to synthesize from all possible candidates is a complex decision-making process that typically relies on expert chemist intuition. We propose a quantitative decision-making framework, SPARROW, that prioritizes molecules for evaluation by balancing expected information gain and synthetic cost. SPARROW integrates molecular design, property prediction, and retrosynthetic planning to balance the utility of testing a molecule with the cost of batch synthesis. We demonstrate through three case studies that the developed algorithm captures the non-additive costs inherent to batch synthesis, leverages common reaction steps and intermediates, and scales to hundreds of molecules. SPARROW is open source and can be found at ***/coleygroup/sparrow. Copyright © 2023, The Authors. All rights reserved.

关键词： Molecules

Evaluating the Impacts of Swapping on the US Decennial Census

学校读者我要写书评

暂无评论

arXiv 2025年

作者： Ballesteros, María Dwork, Cynthia King, Gary Olson, Conlan Raghavan, Manish Institute for Quantitative Social Science Harvard University United States Department of Computer Science Harvard University United States Department of Computer Science Columbia University United States Sloan School of Management Department of Electrical Engineering and Computer Science MIT United States

To meet its dual burdens of providing useful statistics and ensuring privacy of individual respondents, the US Census Bureau has for decades introduced some form of "noise" into published statistics. Initially, they used a method known as "swapping" (1990–2010). In 2020, they switched to an algorithm called TopDown that ensures a form of Differential Privacy. While the TopDown algorithm has been made public, no implementation of swapping has been released and many details of the deployed swapping methodology deployed have been kept secret. Further, the Bureau has not published (even a synthetic) "original" dataset and its swapped version. It is therefore difficult to evaluate the effects of swapping, and to compare these effects to those of other privacy technologies. To address these difficulties we describe and implement a parameterized swapping algorithm based on Census publications, court documents, and informal interviews with Census employees. With this implementation, we characterize the impacts of swapping on a range of statistical quantities of interest. We provide intuition for the types of shifts induced by swapping and compare against those introduced by TopDown. We find that even when swapping and TopDown introduce errors of similar magnitude, the direction in which statistics are biased need not be the same across the two techniques. More broadly, our implementation provides researchers with the tools to analyze and potentially correct for the impacts of disclosure avoidance systems on the quantities they study. Copyright © 2025, The Authors. All rights reserved.

关键词： Differential privacy

A Constructive Proof of the Spherical Parisi Formula

学校读者我要写书评

暂无评论

arXiv 2023年

作者： Huang, Brice Sellke, Mark Department of Electrical Engineering and Computer Science MIT United States Department of Statistics Harvard University United States

The Parisi formula for the free energy is among the crown jewels in the theory of spin glasses. We present a simpler proof of the lower bound in the case of the spherical mean-field model. Our method is constructive: given the minimizer in Parisi's variational problem, we locate a pure state via an explicit optimization algorithm due to Subag on parts with full RSB, and a new truncated second moment argument on critical points for finite RSB parts. Similarly, we show that a truncated second moment suffices to analyze the free energy of a pure state, which amounts to computing the free energy in replica-symmetric models without external field. Our truncation arguments rely on interpolation upper bounds applied to conditional band models. Notably we do not use the Aizenman-Sims-Starr scheme, and require interpolation bounds only up to the 1RSB level. Whenever replica symmetry breaks, we obtain an approximately ultrametric set of exponentially many pure states, each with approximately the same free energy as the entire model, which are hierarchically arranged in accordance with the Parisi ansatz. Our methods also yield results for large deviations of the ground state, including the entire upper tail rate function for all 1RSB models without external field. Copyright © 2023, The Authors. All rights reserved.

关键词： Free energy

Prefix-tree decoding for predicting mass spectra from molecules 23

学校读者我要写书评

暂无评论

Prefix-tree decoding for predicting mass spectra from molecu...

Proceedings of the 37th International Conference on Neural Information Processing Systems

作者： Samuel Goldman John Bradshaw Jiayi Xin Connor W. Coley Computational and Systems Biology MIT Cambridge MA Chemical Engineering MIT Cambridge MA Statistics and Actuarial Science The University of Hong Kong Pokfulam Hong Kong Chemical Engineering Electrical Engineering and Computer Science MIT Cambridge MA

Computational predictions of mass spectra from molecules have enabled the discovery of clinically relevant metabolites. However, such predictive tools are still limited as they occupy one of two extremes, either operating (a) by fragmenting molecules combinatorially with overly rigid constraints on potential rearrangements and poor time complexity or (b) by decoding lossy and nonphysical discretized spectra vectors. In this work, we use a new intermediate strategy for predicting mass spectra from molecules by treating mass spectra as sets of molecular formulae, which are themselves multisets of atoms. After first encoding an input molecular graph, we decode a set of molecular subformulae, each of which specify a predicted peak in the mass spectrum, the intensities of which are predicted by a second model. Our key insight is to overcome the combinatorial possibilities for molecular subformulae by decoding the formula set using a prefix tree structure, atom-type by atom-type, representing a general method for ordered multiset decoding. We show promising empirical results on mass spectra prediction tasks.

关键词：

Combining Convolutional Neural Networks with Reinforcement Learning for Autonomous Robotics

学校读者我要写书评

暂无评论

Combining Convolutional Neural Networks with Reinforcement L...

2025 International Conference on Pervasive Computational Technologies, ICPCT 2025

作者： Padhiary, Mrutyunjay Powar, Swati Asokan, Anju Sundar Rajan, M. Al-Said, Khaleel Al Said, Nidal Assam University Triguna Sen School of Technology Department of Agricultural Engineering Assam Silchar788011 India Mit Adt University Computer Science Engineering Maharashtra Pune India Cse Nehru College of Engineering and Research Centre Kerala Thrissur India Arbaminch University Ethiopia Faculty of Electrical and Computer Engineering Arbaminch Institute of Technology India Middle East University Faculty of Arts and Educational Sciences Department of Educational Technology Amman Jordan Ajman University College of Mass Communication Ajman United Arab Emirates

ISBN: (纸本)9798331508685

The use of artificial intelligence (AI) in robotics, particularly in the advancement of gaming robotics, has drawn a lot of interest as technological and scientific developments continue to advance. This innovation opens up new possibilities for creating smart autonomous robots for gaming. Nevertheless, there is more work to be done to achieve great accuracy and consistency in motion management. The challenges of the robotics' level of smart and the motion-capturing structure's capacity to prevent interruption still require resolution, although numerous methods have been put out to improve monitoring accuracy by merging various data. This study proposes an adaptable reinforcement learning (RL)-oriented multifaceted data combination (AdRL-MDC) system for training a robotic hand to play games alongside humans. It incorporates an adaptable training process for updating the ensemble classification algorithm, an RL paradigm that provides the robots with smart knowledge, and a multifaceted data combination framework that is resistant to interruption. The following studies demonstrate the AdRL-MDC system's previously indicated functionality. The ensemble framework, which combines a convolutional neural network (CNN), performs well in terms of accuracy and computing duration. Furthermore, the depth vision-oriented CNN classification algorithm achieves one hundred percent recognition accuracy, implying that the forecasted motions are the real value. © 2025 IEEE.

关键词： Autonomous robots Convolutional neural network Gaming Motion management Reinforcement Learning

Improving Small Language Models on PubMedQA via Generative Data Augmentation

学校读者我要写书评

暂无评论

arXiv 2023年

作者： Guo, Zhen Wang, Yanwei Wang, Peiqi Yu, Shangdi MIT Department of Electrical Engineering and Computer Science 77 Massachusetts Ave CambridgeMA02139 United States

Large Language Models (LLMs) have made remarkable advancements in the field of natural language processing. However, their increasing size poses challenges in terms of computational cost. On the other hand, Small Language Models (SLMs) are known for their efficiency, but they often struggle with limited capacity and training data, especially in specific domains. In this paper, we introduce a novel method aimed at improving SLMs in the medical domain using LLM-based generative data augmentation. The objective of our approach is to develop more efficient and capable models that are specifically tailored for specialized applications. Through experiments conducted on the PubMedQA dataset, we demonstrate the effectiveness of LLMs in refining and diversifying existing question-answer pairs. This refinement process leads to improved performance in a significantly smaller model after fine-tuning. Notably, our best SLM, with under 1.6 billion parameters, outperforms the few-shot GPT-4 on the PubMedQA dataset. Our code and generated data are publicly available to facilitate further explorations [1]. © 2023, CC BY.

关键词： Natural language processing systems

GRAND-EDGE: A Universal, Jamming-Resilient Algorithm with Error-and-Erasure Decoding

学校读者我要写书评

暂无评论

GRAND-EDGE: A Universal, Jamming-Resilient Algorithm with Er...

IEEE International Conference on Communications (ICC)

作者： Furkan Ercan Kevin Galligan David Starobinski Muriel Médard Ken R. Duffy Rabia Tugce Yazicigil Department of Electrical and Computer Engineering Boston University Boston MA USA Hamilton Institute Maynooth University Ireland Department of Electrical Engineering and Computer Science MIT Cambridge MA USA

Random jammers that overpower transmitted signals are a practical concern for many wireless communication protocols. As such, wireless receivers must be able to cope with standard channel noise and jamming (intentional or unintentional). To address this challenge, we propose a novel method to augment the resilience of the recent family of universal error-correcting GRAND algorithms. This method, called Erasure Decoding by Gaussian Elimination (EDGE), impacts the syndrome check block and is applicable to any variant of GRAND. We show that the proposed EDGE method naturally reverts to the original syndrome check function in the absence of erasures caused by jamming. We demonstrate this by implementing and evaluating GRAND-EDGE and ORBGRAND-EDGE. Simulation results, using a Random Linear Code (RLC) with a code rate of 105/128, show that the EDGE variants lower both the Block Error Rate (BLER) and the computational complexity by up to five order of magnitude compared to the original GRAND and ORBGRAND algorithms. We further compare ORBGRAND-EDGE to Ordered Statistics Decoding (OSD), and demonstrate an improvement of up to three orders of magnitude in the BLER.

关键词：

An Information-Theoretic Framework for Unifying Active Learning Problems 35

学校读者我要写书评

暂无评论

An Information-Theoretic Framework for Unifying Active Learn...

35th AAAI Conference on Artificial Intelligence, AAAI 2021

作者： Nguyen, Quoc Phong Hsiang Low, Bryan Kian Jaillet, Patrick Dept. of Computer Science National University of Singapore Singapore Singapore Dept. of Electrical Engineering and Computer Science MIT United States

ISBN: (纸本)9781713835974

This paper presents an information-theoretic framework for unifying active learning problems: level set estimation (LSE), Bayesian optimization (BO), and their generalized variant. We first introduce a novel active learning criterion that subsumes an existing LSE algorithm and achieves state-of-the-art performance in LSE problems with a continuous input domain. Then, by exploiting the relationship between LSE and BO, we design a competitive information-theoretic acquisition function for BO that has interesting connections to upper confidence bound and max-value entropy search (MES). The latter connection reveals a drawback of MES which has important implications on not only MES but also on other MES-based acquisition functions. Finally, our unifying information-theoretic framework can be applied to solve a generalized problem of LSE and BO involving multiple level sets in a data-efficient manner. We empirically evaluate the performance of our proposed algorithms using synthetic benchmark functions, a real-world dataset, and in hyperparameter tuning of machine learning models. © 2021, Association for the Advancement of Artificial Intelligence (***). All rights reserved.

关键词： Benchmarking