Reproduce “Training a Probabilistic Graphical Model with Resistive Switching Electronic Synapses” 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
S. Burc Eryilmaz, Emre Neftci, Siddharth Joshi, SangBum Kim, Matthew BrightSky, Hsiang-Lan Lung, Chung Lam, Gert Cauwenberghs and 1 more
A Restricted Boltzmann Machine was experimentally implemented and trained using nanoscale phase change memory as artificial synapses, demonstrating unsupervised learning with a bio-inspired algorithm and achieving approximately 150 times lower energy consumption for synaptic operations than conventional digital systems.
Key insights
- The tight integration of memory and processing in neuromorphic architectures using PCM synapses significantly reduces energy consumption by minimizing data movement.
- Practical device characteristics, such as limited control over gradual conductance changes, cycle-to-cycle variations, and device saturation, impact the long-term learning performance of PCM-based RBMs.
- The inherent nonlinearity of PCM resistance changes can initially offer advantages over linear digital synapses in learning performance for certain tasks and learning rates.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.