Tag: model compression

A new knowledge distillation method compresses large pre-trained models into lightweight detectors that identify vulnerabilities in Python code with Macro-F1 ...

A new lightweight model combines self-attention-enhanced YOLOv5 with 2D image sonar to detect and localize small underwater targets at long ...

A new dual-branch AI framework called TBFuseFormer detects tuberculosis on chest X-rays with 99.39 percent held-out test accuracy after removing ...

A new multimodal transformer framework detects REM sleep behavior disorder with an AUC of 0.95 while running in 28 milliseconds, ...

A knowledge-distilled convolutional network reached 92.64 percent accuracy on an independent breast ultrasound dataset, outperforming chief physicians in a retrospective ...

A reinforcement learning-based dynamic quantization method compresses seismic phase-picking models to as little as 3.34 bits on average while keeping ...

A new distillation framework that adjusts its temperature per language and sample compresses a 560-million-parameter multilingual model into a 66-million-parameter ...

A new survey from the National University of Defense Technology systematically compares distributed parallelism and memory-saving techniques that make large ...

A new study systematically benchmarks five CNN compression techniques across GPUs, CPUs, and noisy data, finding that the best choice ...

A new theoretical framework trains neural networks natively on the Leech lattice, compressing weights to 0.76 bits per parameter while ...