Memorizing without overfitting: Bias, variance, and interpolation in over-parameterized models
October 26, 2020 ยท Declared Dead ยท ๐ Physical Review Research
"No code URL or promise found in abstract"
Evidence collected by the PWNC Scanner
Authors
Jason W. Rocks, Pankaj Mehta
arXiv ID
2010.13933
Category
stat.ML: Machine Learning (Stat)
Cross-listed
cond-mat.dis-nn,
cs.LG
Citations
57
Venue
Physical Review Research
Last Checked
5 months ago
Abstract
The bias-variance trade-off is a central concept in supervised learning. In classical statistics, increasing the complexity of a model (e.g., number of parameters) reduces bias but also increases variance. Until recently, it was commonly believed that optimal performance is achieved at intermediate model complexities which strike a balance between bias and variance. Modern Deep Learning methods flout this dogma, achieving state-of-the-art performance using "over-parameterized models" where the number of fit parameters is large enough to perfectly fit the training data. As a result, understanding bias and variance in over-parameterized models has emerged as a fundamental problem in machine learning. Here, we use methods from statistical physics to derive analytic expressions for bias and variance in two minimal models of over-parameterization (linear regression and two-layer neural networks with nonlinear data distributions), allowing us to disentangle properties stemming from the model architecture and random sampling of data. In both models, increasing the number of fit parameters leads to a phase transition where the training error goes to zero and the test error diverges as a result of the variance (while the bias remains finite). Beyond this threshold, the test error of the two-layer neural network decreases due to a monotonic decrease in \emph{both} the bias and variance in contrast with the classical bias-variance trade-off. We also show that in contrast with classical intuition, over-parameterized models can overfit even in the absence of noise and exhibit bias even if the student and teacher models match. We synthesize these results to construct a holistic understanding of generalization error and the bias-variance trade-off in over-parameterized models and relate our results to random matrix theory.
Community Contributions
Found the code? Know the venue? Think something is wrong? Let us know!
๐ Similar Papers
In the same crypt โ Machine Learning (Stat)
๐ฎ
๐ฎ
The Ethereal
๐ฎ
๐ฎ
The Ethereal
Layer Normalization
๐ฎ
๐ฎ
The Ethereal
Simple and Scalable Predictive Uncertainty Estimation using Deep Ensembles
R.I.P.
๐ป
Ghosted
Variational Inference with Normalizing Flows
๐
๐
The Cartographer
Towards A Rigorous Science of Interpretable Machine Learning
R.I.P.
๐ป
Ghosted
Optimization Methods for Large-Scale Machine Learning
Died the same way โ ๐ป Ghosted
R.I.P.
๐ป
Ghosted
Federated Learning: Strategies for Improving Communication Efficiency
R.I.P.
๐ป
Ghosted
In-Datacenter Performance Analysis of a Tensor Processing Unit
R.I.P.
๐ป
Ghosted
Deep Convolutional Neural Networks for Computer-Aided Detection: CNN Architectures, Dataset Characteristics and Transfer Learning
R.I.P.
๐ป
Ghosted