AI & ML
ASDL: A Fast Vector Approximation Algorithm for Multiplierless Designs of Convolutional Neural Networks
arXiv eess.SP0 views
arXiv:2611.21103v2 Announce Type: replace Abstract: Matrix multiplications are one of the main computational bottlenecks in neural networks. The calculation reduces to numerous sums of partial products, which scale with the total number of Signed Power of Two (SPT) values in the multiplier matrix. Minimizing the SPT count can significantly reduce the computational burden. In this context, we present a Sum Of (signed) Powers Of Two (SOPOT) vector approximation algorithm that limits SPT counts, namely the Accelerated Signed Digit Loading (ASDL) algorithm. This approximation algorithm converts ResNet50 10 times faster than its predecessor MPGBP, while being more precise. We apply this method to the pre-trained coefficients of different convolutional neural networks, achieving 80.41\% ImageNet accuracy down from 81.3\%, at an average of only 1.5 active bits per coefficient. This is done without any retraining or calibration, which are typically required for the most aggressive quantization methods. Our approach offers a promising direction for neural network inference with higher energy efficiency and fewer hardware requirements: we anticipate up to 10 times improvement over bfloat16 inference.
Source:https://arxiv.org/abs/2611.21103PreviousLearning-Driven Channel Representation for Wireless Localization: From Channel Observations to Location InferenceNextSex Differences in the Detection of Parkinson's Disease from Speech
Related posts
Streaming Generative Video Compression with Embedding Alignment and In-Context Propagation
0arXiv eess.IV
Beyond Accuracy: Benchmarking Uncertainty Quantification in Retinal Disease Screening
0arXiv eess.IV
CiGS: Conduction in Gaussian Splats for Interior Temperature Field from Thermal Images
0arXiv cs.GR
Probing Physics in Motion Diffusion Models Against Their Own Estimate
0arXiv cs.GR
NIRJALA: Draining the Sea out of Gaussian Splatting with Three-Path Optics and Reversible Physics
0arXiv cs.GR
How Music LLMs Actually Work: From Tokens to Waves (MusicGen, DiT, and Building Your Own Engine)
0dev.to Machine Learning