Reproduce “SparseForge: Beyond Spike Sparsity toward Hardware-Exploitable Neuron Pruning for Efficient SNN Acceleration” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Kainan Wang, Sanhe Fu, Aili Wang
SparseForge co-designs spiking neural network inference by learning fixed neuron masks and mapping them to sparse FPGA dataflow. On VGGSNN for DVS128-Gesture at 20 time steps, Low pruning retains 6.76% of neurons and 1.51% of synaptic operations, while accuracy rises from 96.88% to 97.57%. This improvement is specific to that task and setting.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.