Sparsely ensembled convolutional neural network classifiers via reinforcement learning

TL;DR

This paper introduces a reinforcement learning-based method for dynamically selecting and combining CNN classifiers to optimize accuracy and computational efficiency in ensemble learning.

Contribution

It proposes a novel reinforcement learning approach to dynamically configure CNN ensembles, improving efficiency and accuracy over traditional static ensembles.

Findings

Reinforcement learning agent effectively learns classifier selection strategies.

Dynamic ensemble configuration outperforms conventional static ensembles.

The method reduces computational resource consumption while maintaining high accuracy.

Abstract

We consider convolutional neural network (CNN) ensemble learning with the objective function inspired by least action principle; it includes resource consumption component. We teach an agent to perceive images through the set of pre-trained classifiers and want the resulting dynamically configured system to unfold the computational graph with the trajectory that refers to the minimal number of operations and maximal expected accuracy. The proposed agent's architecture implicitly approximates the required classifier selection function with the help of reinforcement learning. Our experimental results prove, that if the agent exploits the dynamic (and context-dependent) structure of computations, it outperforms conventional ensemble learning.