检索结果-内蒙古大学图书馆

Variational Bayesian unlearning 34

学校读者我要写书评

暂无评论

Variational Bayesian unlearning

34th Conference on Neural Information Processing Systems, NeurIPS 2020

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

This paper studies the problem of approximately unlearning a Bayesian model from a small subset of the training data to be erased. We frame this problem as one of minimizing the Kullback-Leibler divergence between the approximate posterior belief of model parameters after directly unlearning from erased data vs. the exact posterior belief from retraining with remaining data. Using the variational inference (VI) framework, we show that it is equivalent to minimizing an evidence upper bound which trades off between fully unlearning from erased data vs. not entirely forgetting the posterior belief given the full data (i.e., including the remaining data);the latter prevents catastrophic unlearning that can render the model useless. In model training with VI, only an approximate (instead of exact) posterior belief given the full data can be obtained, which makes unlearning even more challenging. We propose two novel tricks to tackle this challenge. We empirically demonstrate our unlearning methods on Bayesian models such as sparse Gaussian process and logistic regression using synthetic and real-world datasets. © 2020 Neural information processing systems foundation. All rights reserved.

关键词：

LEDRO: LLM-Enhanced Design Space Reduction and Optimization for Analog Circuits

学校读者我要写书评

暂无评论

arXiv 2024年

作者： Kochar, Dimple Vijay Wang, Hanrui Chandrakasan, Anantha P. Zhang, Xin Electrical Engineering and Computer Science Massachusetts Institute of Technology CambridgeMA United States MIT-IBM Watson AI Lab CambridgeMA United States IBM T. J. Watson Research Center Yorktown HeightsNY United States

Traditional approaches for designing analog circuits are time-consuming and require significant human expertise. Existing automation efforts using methods like Bayesian Optimization (BO) and Reinforcement Learning (RL) are suboptimal and costly to generalize across different topologies and technology nodes. In our work, we introduce a novel approach, LEDRO, utilizing Large Language Models (LLMs) in conjunction with optimization techniques to iteratively refine the design space for analog circuit sizing. LEDRO is highly generalizable compared to other RL and BO baselines, eliminating the need for design annotation or model training for different topologies or technology nodes. We conduct a comprehensive evaluation of our proposed framework and baseline on 22 different Op-Amp topologies across four FinFET technology nodes. Results demonstrate the superior performance of LEDRO as it outperforms our best baseline by an average of 13% FoM improvement with 2.15× speed-up on low-complexity Op-Amps and 48% FoM improvement with 1.7× speed-up on high-complexity Op-Amps. This highlights LEDRO’s effective performance, efficiency, and generalizability. Copyright © 2024, The Authors. All rights reserved.

关键词： Operational amplifiers

Assessing Affective Engagement with Narratives of Invisible Disability

学校读者我要写书评

暂无评论

Assessing Affective Engagement with Narratives of Invisible ...

International Conference on Affective Computing and Intelligent Interaction and Workshops, ACII

作者： Daniel T. Kessler David Y.J. Kim Grace S. Ahn Neska Elhaouij Rosalind Picard MIT Center for Constructive Communication Massachusetts Institute of Technology Cambridge USA Electrical Engineering and Computer Science Massachusetts Institute of Technology Cambridge USA MIT System Design and Management Massachusetts Institute of Technology Cambridge USA Affective Computing Group Massachusetts Institute of Technology Cambridge USA

Narratives about invisible disabilities are poorly represented in public discourse and often go undisclosed [1], leading to false assumptions, discrimination, and stigma [2] against those who experience these conditions. To address these issues, recent studies have suggested that disclosure of first-person narratives of invisible disabilities should be increased [3]. To understand the mechanisms affecting recipients of such narratives, the present study evaluates how social media users (N = 124) engage affectively with this content in a digitally mediated narrative-form intervention designed to reduce harmful assumptions against persons who experience invisible disabilities. Results of this study indicate that such an intervention may prove effective at reducing harmful assumptions on the basis of visual cues, and in line with past research, finds that affect may play an important role in assumption-making processes [4]. Findings from this study may be used to inform novel digital interventions capable of counteracting harmful assumptions that drive prejudicial behaviors against a wide range of populations and communities.

关键词：

MetaMorph: Learning Metamorphic Image Transformation With Appearance Changes

学校读者我要写书评

暂无评论

arXiv 2023年

作者： Wang, Jian Xing, Jiarui Druzgal, Jason Wells, William M. Zhang, Miaomiao Computer Science University of Virginia VA United States Electrical and Computer Engineering University of Virginia VA United States Radiology and Medical Imaging University of Virginia VA United States Brigham and Women’s Hospital Harvard Medical School MA United States Computer Science and Artificial Intelligence Laboratory MIT MA United States

This paper presents a novel predictive model, MetaMorph, for metamorphic registration of images with appearance changes (i.e., caused by brain tumors). In contrast to previous learning-based registration methods that have little or no control over appearance-changes, our model introduces a new regularization that can effectively suppress the negative effects of appearance changing areas. In particular, we develop a piecewise regularization on the tangent space of diffeomorphic transformations (also known as initial velocity fields) via learned segmentation maps of abnormal regions. The geometric transformation and appearance changes are treated as joint tasks that are mutually beneficial. Our model MetaMorph is more robust and accurate when searching for an optimal registration solution under the guidance of segmentation, which in turn improves the segmentation performance by providing appropriately augmented training labels. We validate MetaMorph on real 3D human brain tumor magnetic resonance imaging (MRI) scans. Experimental results show that our model outperforms the state-of-the-art learning-based registration models. The proposed MetaMorph has great potential in various image-guided clinical interventions, e.g., real-time image-guided navigation systems for tumor removal surgery. Copyright © 2023, The Authors. All rights reserved.

关键词： Tumors

Deformed2Self: Self-supervised denoising for dynamic medical imaging

学校读者我要写书评

暂无评论

arXiv 2021年

作者： Xu, Junshen Adalsteinsson, Elfar Department of Electrical Engineering and Computer Science Mit CambridgeMA United States Institute for Medical Engineering and Science Mit CambridgeMA United States

Image denoising is of great importance for medical imaging system, since it can improve image quality for disease diagnosis and downstream image analyses. In a variety of applications, dynamic imaging techniques are utilized to capture the time-varying features of the subject, where multiple images are acquired for the same subject at different time points. Although signal-to-noise ratio of each time frame is usually limited by the short acquisition time, the correlation among different time frames can be exploited to improve denoising results with shared information across time frames. With the success of neural networks in computer vision, supervised deep learning methods show prominent performance in single-image denoising, which rely on large datasets with clean-vs-noisy image pairs. Recently, several self-supervised deep denoising models have been proposed, achieving promising results without needing the pairwise ground truth of clean images. In the field of multi-image denoising, however, very few works have been done on extracting correlated information from multiple slices for denoising using self-supervised deep learning methods. In this work, we propose Deformed2Self, an end-to-end self-supervised deep learning framework for dynamic imaging denoising. It combines single-image and multi-image denoising to improve image quality and use a spatial transformer network to model motion between different slices. Further, it only requires a single noisy image with a few auxiliary observations at different time frames for training and inference. Evaluations on phantom and in vivo data with diferent noise statistics show that our method has comparable performance to other state-of-the-art unsupervised or self-supervised denoising methods and outperforms under high noise levels. Copyright © 2021, The Authors. All rights reserved.

关键词： Image denoising

Machine learning's own Industrial Revolution

学校读者我要写书评

暂无评论

arXiv 2023年

作者： Luo, Yuan Han, Song Liu, Jingjing Department of Preventive Medicine Northwestern University ChicagoIL60611 United States Electrical Engineering and Computer Science MIT CambridgeMA02139 United States Institute for AI Industry Research Tsinghua University Beijing China

Machine learning is expected to enable the next Industrial Revolution. However, lacking standardized and automated assembly networks, ML faces significant challenges to meet ever-growing enterprise demands and empower broad industries. In the Perspective, we argue that ML needs to first complete its own Industrial Revolution, elaborate on how to best achieve its goals, and discuss new opportunities to enable rapid translation from ML's innovation frontier to mass production and utilization. © 2023, CC BY.

关键词： Machine learning

High-Fidelity, Frequency-Flexible Two-Qubit Fluxonium Gates with a Transmon Coupler

学校读者我要写书评

暂无评论

Physical Review X 2023年第3期13卷 031035-031035页

作者： Leon Ding Max Hays Youngkyu Sung Bharath Kannan Junyoung An Agustin Di Paolo Amir H. Karamlou Thomas M. Hazard Kate Azar David K. Kim Bethany M. Niedzielski Alexander Melville Mollie E. Schwartz Jonilyn L. Yoder Terry P. Orlando Simon Gustavsson Jeffrey A. Grover Kyle Serniak William D. Oliver Department of Physics Massachusetts Institute of Technology Cambridge Massachusetts 02139 USA Research Laboratory of Electronics Massachusetts Institute of Technology Cambridge Massachusetts 02139 USA Department of Electrical Engineering and Computer Science Massachusetts Institute of Technology Cambridge Massachusetts 02139 USA MIT Lincoln Laboratory Lexington Massachusetts 02421 USA

We propose and demonstrate an architecture for fluxonium-fluxonium two-qubit gates mediated by transmon couplers (FTF, for fluxonium-transmon-fluxonium). Relative to architectures that exclusively rely on a direct coupling between fluxonium qubits, FTF enables stronger couplings for gates using noncomputational states while simultaneously suppressing the static controlled-phase entangling rate (ZZ) down to kilohertz levels, all without requiring strict parameter matching. Here, we implement FTF with a flux-tunable transmon coupler and demonstrate a microwave-activated controlled-Z (CZ) gate whose operation frequency can be tuned over a 2-GHz range, adding frequency allocation freedom for FTFs in larger systems. Across this range, state-of-the-art CZ gate fidelities are observed over many bias points and reproduced across the two devices characterized in this work. After optimizing both the operation frequency and the gate duration, we achieve peak CZ fidelities in the 99.85%–99.9% range. Finally, we implement model-free reinforcement learning of the pulse parameters to boost the mean gate fidelity up to 99.922%±0.009%, averaged over roughly an hour between scheduled training runs. Beyond the microwave-activated CZ gate we present here, FTF can be applied to a variety of other fluxonium gate schemes to improve gate fidelities and passively reduce unwanted ZZ interactions.

关键词： Machine learning Quantum circuits Quantum coherence & coherence measures Quantum control Quantum engineering Quantum gates Quantum information architectures & platforms

Patch-level Routing in Mixture-of-Experts is Provably Sample-efficient for Convolutional Neural Networks

学校读者我要写书评

暂无评论

arXiv 2023年

作者： Chowdhury, Mohammed Nowaz Rabbani Zhang, Shuai Wang, Meng Liu, Sijia Chen, Pin-Yu Department of Electrical Computer and Systems Engineering Rensselaer Polytechnic Institute NY United States Department of Computer Science and Engineering Michigan State University MI United States MIT-IBM Watson Ai Lab Ibm Research MA United States Ibm Research Yorktown HeightsNY United States

In deep learning, mixture-of-experts (MoE) activates one or few experts (sub-networks) on a per-sample or per-token basis, resulting in significant computation reduction. The recently proposed patch-level routing in MoE (pMoE) divides each input into n patches (or tokens) and sends l patches (l n) to each expert through prioritized routing. pMoE has demonstrated great empirical success in reducing training and inference costs while maintaining test accuracy. However, the theoretical explanation of pMoE and the general MoE remains elusive. Focusing on a supervised classification task using a mixture of two-layer convolutional neural networks (CNNs), we show for the first time that pMoE provably reduces the required number of training samples to achieve desirable generalization (referred to as the sample complexity) by a factor in the polynomial order of n/l, and outperforms its single-expert counterpart of the same or even larger capacity. The advantage results from the discriminative routing property, which is justified in both theory and practice that pMoE routers can filter label-irrelevant patches and route similar class-discriminative patches to the same expert. Our experimental results on MNIST, CIFAR-10, and CelebA support our theoretical findings on pMoE's generalization and show that pMoE can avoid learning spurious correlations. Copyright © 2023, The Authors. All rights reserved.

关键词： Convolution

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...

Pervasive Computational Technologies (ICPCT), International Conference on

作者： Mrutyunjay Padhiary Swati Powar Anju Asokan M. Sundar Rajan Khaleel Al-Said Nidal Al Said Department of Agricultural Engineering Triguna Sen School of Technology Assam University Silchar Assam Computer Science Engineering MIT ADT University Pune Maharashtra CSE Nehru College of Engineering and Research Centre Thrissur Kerala Faculty of Electrical and Computer Engineering Arbaminch Institute of Technology Arbaminch University Ethiopia Department of Educational Technology Faculty of Arts and Educational Sciences Middle East University Amman Jordan College of Mass Communication Ajman University Ajman UAE

ISBN: (数字)9798331508685

ISBN: (纸本)9798331519476

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.

关键词： Training Resistance Technological innovation Accuracy Merging Reinforcement learning Classification algorithms Convolutional neural networks Monitoring Autonomous robots