Structured Neuron Pruning in Deep Neural Networks Using Multi-Armed Bandits

In three linesStructured pruning framework using multi-armed bandit (MAB) algorithms to remove complete neurons from deep neural networks. Evaluates UCB1, Thompson Sampling, Epsilon-Greedy and other policies on classification, regression and deep learning tasks. UCB1 and Thompson Sampling outperform magnitude-based pruning and unpruned models.

Summary generated by Claude — human-verified