Random Cloud: Finding Minimal Neural Architectures Without Training

TL;DR

The paper introduces Random Cloud, a training-free neural architecture search method that efficiently finds minimal network topologies by stochastic exploration and structural reduction, outperforming traditional pruning methods on several benchmarks.

Contribution

It presents a novel training-free approach to neural architecture search that reduces training costs and improves performance over baseline pruning methods.

Findings

Achieves better or comparable accuracy to pruning baselines on 6 of 7 datasets.

Reduces parameters by up to 87% while maintaining performance.

Faster than pruning baselines in most cases by avoiding full training.

Abstract

I propose the \emph{Random Cloud} method, a training-free approach to neural architecture search that discovers minimal feedforward network topologies through stochastic exploration and progressive structural reduction. Unlike post-training pruning methods that require a full train-prune-retrain cycle, this method evaluates randomly initialized networks without backpropagation, progressively reduces their topology, and only trains the best minimal candidate at the end. I evaluate on 7 classification benchmarks against magnitude pruning and random pruning baselines. The Random Cloud matches or outperforms both baselines in 6 of 7 datasets, achieving statistically significant improvements on Sonar ($+4.9$pp accuracy, $p{=}0.017$ vs magnitude pruning) with 87\% parameter reduction. Crucially, the method is faster than both pruning baselines in 4 of 5 datasets (0.67--0.94$\times$ the cost of full training), since it avoids training the full-size network entirely.