Purdue's SQS Combines Spike-and-Slab Sparsity and GMM Quantization for High-Rate DNN Compression
Purdue · hf · 2026-09-09
Purdue researchers released SQS, a unified Bayesian variational framework for compressing large neural networks.
- Combines spike-and-slab sparsity with Gaussian mixture quantization in one framework
- Achieves high compression rates with minimal accuracy loss
- Offers a principled Bayesian approach unifying pruning and quantization
More from Research
- Meta researcher's training trick: drop a batch proportion instead of per-sample to keep GPU efficiency — TimDarcet · 2026-09-09
- Common Crawl releases experimental graph embeddings for 52.9M web hosts — lhoestq · 2026-09-09
- 3DHarnessBench probes agentic 3D-to-code skills of frontier VLMs — ftm_guney · 2026-09-09
- Anthropic researcher: novelty-rewarded RL may teach agents to obfuscate their sources — suchenzang · 2026-09-09
- Why swarm trajectories may be the key missing training signal for research RL — teortaxesTex · 2026-09-09
- OpenAI's Navier-Stokes run used a model just 4 days into training, still unfinished — haider1 · 2026-09-09