Reproduce “Accelerating Deep Learning with Memcomputing” 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
Haik Manukian, Fabio L. Traversa, Massimiliano Di Ventra
The authors introduce Mem-QUBO, using a simulated memcomputing machine to search low-energy states in restricted Boltzmann machines during unsupervised digit pretraining. It replaces contrastive divergence’s short Gibbs-sampling runs with an approximate optimization of the model expectation, without guaranteeing a global optimum. On reduced MNIST, it reaches accuracies in fewer pretraining iterations than CD-1; comparison with a batch-normalized ReLU network does not isolate pretraining.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.