paper-with-me

Papers

A Bayesian Perspective on Training Speed and Model Selection

2020-10-27 · NeurIPS 2020 12 · Clare Lyle, Lisa Schut, Binxin Ru, Yarin Gal, Mark van der Wilk

We take a Bayesian perspective to illustrate a connection between training speed and the marginal likelihood in linear models. This provides two major insights: first, that a measure of a model's training speed can be used to estimate its marginal likelihood. Second, that this measure, under certain conditions, predicts the relative weighting of models in linear model combinations trained to minimize a regression loss. We verify our results in model selection tasks for linear models and for the infinite-width limit of deep neural networks. We further provide encouraging empirical evidence that the intuition developed in these settings also holds for deep neural networks trained with stochastic gradient descent. Our results suggest a promising new direction towards explaining why neural networks trained with stochastic gradient descent are biased towards functions that generalize well.

📄 PDF Abstract BibTeX arXiv:2010.14499

Code (0)

등록된 구현이 없습니다.

Tasks

Model Selection

Similar Papers 제목 키워드 기반

Bayesian Coresets: Revisiting the Nonconvex Optimization Perspective

2020-07-01 · Jacky Y. Zhang, Rajiv Khanna, Anastasios Kyrillidis, Oluwasanmi Koyejo

Bayesian coresets have emerged as a promising approach for implementing scalable Bayesian inference. The Bayesian coreset problem involves selecting a (weighted) subset of the data samples, such that the posterior infere…

Bayesian Inference

Bayesian subset selection and variable importance for interpretable prediction and classification

2021-04-20 · Daniel R. Kowal

Subset selection is a valuable tool for interpretable learning, scientific discovery, and data compression. However, classical subset selection is often avoided due to selection instability, lack of regularization, and d…

Data CompressionGeneral Classificationscientific discoveryUncertainty Quantification+1

Bayesian Quantile Regression with Subset Selection: A Decision Analysis Perspective

2023-11-03 · Joseph Feldman, Daniel Kowal

Quantile regression is a powerful tool for inferring how covariates affect specific percentiles of the response distribution. Existing methods either estimate conditional quantiles separately for each quantile of interes…

Density Estimationquantile regressionregressionUncertainty Quantification+1

Deep Learning: A Bayesian Perspective

2017-06-01 · Nicholas Polson, Vadim Sokolov

Deep learning is a form of machine learning for nonlinear high dimensional pattern matching and prediction. By taking a Bayesian probabilistic perspective, we provide a number of insights into more efficient algorithms f…

Deep LearningregressionVariable Selection

Enhancing Offline Model-Based RL via Active Model Selection: A Bayesian Optimization Perspective

2025-02-17 · Yu-Wei Yang, Yun-Ming Chan, Wei Hung, Xi Liu 외

Offline model-based reinforcement learning (MBRL) serves as a competitive framework that can learn well-performing policies solely from pre-collected data with the help of learned dynamics models. To fully unleash the po…

Bayesian OptimizationmodelModel-based Reinforcement LearningModel Selection+2