AdaNet: Adaptive Structural Learning of Artificial Neural Networks

TL;DR

AdaNet introduces algorithms that adaptively learn neural network structures and weights, providing theoretical guarantees and demonstrating competitive performance on image classification tasks.

Contribution

The paper presents novel algorithms for adaptive neural network structure learning with proven generalization guarantees, advancing automatic model design.

Findings

Achieved competitive accuracy on CIFAR-10 classification tasks

Demonstrated effective automatic structure learning for neural networks

Provided theoretical analysis with data-dependent guarantees

Abstract

We present new algorithms for adaptively learning artificial neural networks. Our algorithms (AdaNet) adaptively learn both the structure of the network and its weights. They are based on a solid theoretical analysis, including data-dependent generalization guarantees that we prove and discuss in detail. We report the results of large-scale experiments with one of our algorithms on several binary classification tasks extracted from the CIFAR-10 dataset. The results demonstrate that our algorithm can automatically learn network structures with very competitive performance accuracies when compared with those achieved for neural networks found by standard approaches.