AGENTIC-SYSTEMSJuly 28, 20261 min read

Loss-Aware Feature-Map Pruning in Convolutional Neural Networks Using Multi-Armed Bandits

Researchers have proposed a loss-aware structured feature-map pruning framework for convolutional neural networks using multi-armed bandits. The approach significantly reduces computational complexity and storage costs while maintaining accuracy levels close to original unpruned models.

Computer Science > Artificial Intelligence

Title: Loss-Aware Feature-Map Pruning in Convolutional Neural Networks Using Multi-Armed Bandits

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.

Source: arXiv cs.AI Recent

More in this category

agentic-systems

Schema-Aware Localisation (SAL): Live Schema Grounding and Hallucination Validation for Oracle NL2SQL

Researchers introduce Schema-Aware Localisation (SAL), a lightweight middleware for Oracle NL2SQL that connects LLMs to live database catalogs. SAL dramatically improves SQL execution accuracy from 2.2% to 62.6% while drastically reducing hallucination-induced execution failures without requiring model retraining.

agentic-systems

SCAIR: Schema-Conditioned Agentic Iterative Reasoning for Enterprise Knowledge Graphs

Researchers have introduced SCAIR, a training-free agentic framework that enhances Knowledge Graph-based RAG (KG-RAG) for enterprise applications by injecting schema constraints into multi-hop iterative reasoning.

agentic-systems

Synthetic Scenario Generation for Evaluation of Industry 4.0 Agents

Researchers have developed ScenarioGeneratorAgent, an automated pipeline that generates synthetic evaluation scenarios for Industry 4.0 AI agents. Tested on Smart Grid Transformers, the approach accelerates benchmark generation by 8x while maintaining high quality and adherence to IEC standards.

agentic-systems

Concept-based Visual Counterfactual Explanations with Diffusion Models

Researchers introduced C-VCE, a novel diffusion framework that generates human-interpretable visual counterfactual explanations by building concept bottleneck layers directly into the generative model.

agentic-systems

Teaching LLMs to Update Beliefs for Efficient Long-Horizon Interaction

As LLMs tackle longer tasks, retaining complete interaction histories becomes contextually and computationally expensive. The ABBEL framework addresses this by isolating summaries into natural-language 'belief states' and supervising them via belief grading, recovering accuracy while significantly cutting memory usage.

agentic-systems

DC-Leap: Training-Free Acceleration of dLLMs via Draft-Guided Contiguous Leaping Decoding

Researchers have introduced DC-Leap, a training-free framework that accelerates inference in Diffusion Large Language Models (dLLMs) by up to 105x when combined with KV-Cache while maintaining generation quality.