Fisher Task Distance and Its Application in Neural Architecture Search

Abstract

We formulate an asymmetric (or non-commutative) distance between tasks based on Fisher Information Matrices, called Fisher task distance. This distance represents the complexity of transferring the knowledge from one task to another. We provide a proof of consistency for our distance through theorems and experiments on various classification tasks from MNIST, CIFAR-10, CIFAR-100, ImageNet, and Taskonomy datasets. Next, we construct an online neural architecture search framework using the Fisher task distance, in which we have access to the past learned tasks. By using the Fisher task distance, we can identify the closest learned tasks to the target task, and utilize the knowledge learned from these related tasks for the target task. Here, we show how the proposed distance between a target task and a set of learned tasks can be used to reduce the neural architecture search space for the target task. The complexity reduction in search space for task-specific architectures is achieved by building on the optimized architectures for similar tasks instead of doing a full search and without using this side information. Experimental results for tasks in MNIST, CIFAR-10, CIFAR-100, ImageNet datasets demonstrate the efficacy of the proposed approach and its improvements, in terms of the performance and the number of parameters, over other gradient-based search methods, such as ENAS, DARTS, PC-DARTS.

Published in IEEE Access, Volume 10, 2022

References in corpus (12)

• Very Deep Convolutional Networks for Large-Scale Image Recognition
• Neural Architecture Search with Reinforcement Learning
• PathNet: Evolution Channels Gradient Descent in Super Neural Networks
• PC-DARTS: Partial Channel Connections for Memory-Efficient Architecture Search
• Which Tasks Should Be Learned Together in Multi-task Learning?
• Evaluating the Search Phase of Neural Architecture Search
• ASAP: Architecture Search, Anneal and Prune
• Optimal non-asymptotic bound of the Ruppert-Polyak averaging without strong convexity
• Differential Evolution for Neural Architecture Search
• One-Shot Neural Architecture Search via Compressive Sensing
• Optimal Transport Kernels for Sequential and Parallel Neural Architecture Search
• Efficient Forward Architecture Search