Continuously Constructive Deep Neural Networks

TL;DR

This paper introduces two innovative methods that automatically adapt neural network architecture during training by continuously adjusting complexity, demonstrated on synthetic and real datasets like MNIST and MIRFLICKR.

Contribution

It presents two novel approaches for dynamic neural network architecture adjustment through continuous parameterization, enabling automatic complexity tuning during training.

Findings

Methods effectively adjust network complexity to task difficulty.

Successful application on synthetic and real datasets.

Achieves correct complexity without hyperparameter tuning.

Abstract

Traditionally, deep learning algorithms update the network weights whereas the network architecture is chosen manually, using a process of trial and error. In this work, we propose two novel approaches that automatically update the network structure while also learning its weights. The novelty of our approach lies in our parameterization where the depth, or additional complexity, is encapsulated continuously in the parameter space through control parameters that add additional complexity. We propose two methods: In tunnel networks, this selection is done at the level of a hidden unit, and in budding perceptrons, this is done at the level of a network layer; updating this control parameter introduces either another hidden unit or another hidden layer. We show the effectiveness of our methods on the synthetic two-spirals data and on two real data sets of MNIST and MIRFLICKR, where we see that our proposed methods, with the same set of hyperparameters, can correctly adjust the network complexity to the task complexity.