Predicting Neural Network Accuracy from Weights

by   Thomas Unterthiner, et al.

We study the prediction of the accuracy of a neural network given only its weights with the goal of better understanding network training and performance. To do so, we propose a formal setting which frames this task and connects to previous work in this area. We collect (and release) a large dataset of almost 80k convolutional neural networks trained on four image datasets. We demonstrate that strong predictors of accuracy exist. Moreover, they can achieve good predictions while only using simple statistics of the weights. Surprisingly, these predictors are able to rank networks trained on unobserved datasets or using different architectures.


page 15

page 17


What's Hidden in a Randomly Weighted Neural Network?

Training a neural network is synonymous with learning the values of the ...

Growing Artificial Neural Networks

Pruning is a legitimate method for reducing the size of a neural network...

A Hybrid Method for Training Convolutional Neural Networks

Artificial Intelligence algorithms have been steadily increasing in popu...

Deep Convolutional Neural Networks with Unitary Weights

While normalizations aim to fix the exploding and vanishing gradient pro...

Neural Networks Assist Crowd Predictions in Discerning the Veracity of Emotional Expressions

Crowd predictions have demonstrated powerful performance in predicting f...

Benchmarking Neural Network Generalization for Grammar Induction

How well do neural networks generalize? Even for grammar induction tasks...

Epistemic Modeling Uncertainty of Rapid Neural Network Ensembles for Adaptive Learning

Emulator embedded neural networks, which are a type of physics informed ...

Code Repositories


Features extractors for tensorflow models used to estimate generalization gap

view repo

Please sign up or login with your details

Forgot password? Click here to reset