aiqbench analysis
This research introduces a loss-aware pruning framework for convolutional neural networks that leverages multi-armed bandits to efficiently identify redundant feature maps without compromising model accuracy. By treating channel removal as an optimization problem, the method statistically outperforms traditional greedy or magnitude-based approaches while significantly reducing inference costs on standard benchmarks like CIFAR and ImageNet subsets. The findings offer a practical pathway for deploying more efficient deep learning models in resource-constrained environments where structured sparsity is critical.
Loss-Aware Feature-Map Pruning in Convolutional Neural Networks Using Multi-Armed Bandits
arXiv:2607.22564v1 Announce Type: new Abstract: Convolutional neural networks often contain redundant feature maps that increase storage and inference cost. This paper presents a loss-aware feature-map pruning framework using multi-armed bandits. Feature-map pruning is structured because it removes complete convolutional output channels and their producing filters rather than isolated scalar weights. Each candidate feature map is treated as an arm. At each play time, one map is temporarily masked and evaluated on a sampled mini-batch; the map is then restored and the observed loss change is converted into a safe-removal reward. After a fixed play budget, candidate maps are ranked by learned scores and the top-k maps are permanently removed with their filters, biases and corresponding next-layer input-channel kernels. The study evaluates UCB1 and Thompson Sampling, compares them with direct/oracle-style evaluation on LeNet/MNIST, and extends the evaluation to MNIST, CIFAR-10, CIFAR-100, SVHN, CUB-200-2011 and Oxford Flowers 102. Results show that UCB1 and Thompson Sampling preserve accuracy close to unpruned models while removing feature maps and reducing convolutional computation. Friedman and Nemenyi tests show that UCB1 obtains the highest mean rank, followed by Thompson Sampling; both significantly outperform greedy and magnitude-based pruning while remaining statistically comparable to the original unpruned model.
Abstract:Convolutional neural networks often contain redundant feature maps that increase storage and inference cost. This paper presents a loss-aware feature-map pruning framework using multi-armed bandits. Feature-map pruning is structured because it removes complete convolutional output channels and their producing filters rather than isolated scalar weights. Each candidate feature map is treated as an arm. At each play time, one map is temporarily masked and evaluated on a sampled mini-batch; the map is then restored and the observed loss change is converted into a safe-removal reward. After a fixed play budget, candidate maps are ranked by learned scores and the top-k maps are permanently removed with their filters, biases and corresponding next-layer input-channel kernels. The study evaluates UCB1 and Thompson Sampling, compares them with direct/oracle-style evaluation on LeNet/MNIST, and extends the evaluation to MNIST, CIFAR-10, CIFAR-100, SVHN, CUB-200-2011 and Oxford Flowers 102. Results show that UCB1 and Thompson Sampling preserve accuracy close to unpruned models while removing feature maps and reducing convolutional computation. Friedman and Nemenyi tests show that UCB1 obtains the highest mean rank, followed by Thompson Sampling; both significantly outperform greedy and magnitude-based pruning while remaining statistically comparable to the original unpruned model.
Artificial Intelligence (cs.AI)
arXiv:2607.22564 [cs.AI]
(or arXiv:2607.22564v1 [cs.AI] for this version)
https://doi.org/10.48550/arXiv.2607.22564 Focus to learn more arXiv-issued DOI via DataCite
arXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website.
Both individuals and organizations that work with arXivLabs have embraced and accepted our values of openness, community, excellence, and user data privacy. arXiv is committed to these values and only works with partners that adhere to them.
Have an idea for a project that will add value for arXiv's community? Learn more about arXivLabs.
Full text under the publisher's open licence · 286 words
Related coverage
Build an explainable next-best-product recommendation system for banking on AWS
AWS · Jul 24, 2026
CoG-Guided Weight Correction for Fault-Tolerant Deep Neural Networks
arXiv · Jul 20, 2026
[AINews] High Taste Testers Ep. 1
Latent Space · Oct 17, 2026
PIRAMID: Progress and Plans
LessWrong · Jul 27, 2026