Stochastic Computing
20 days ago
- #Machine Learning
- #Analog Computing
- #Stochastic Computing
- Stochastic computing is an old idea that originated in the 1960s, with early work by Brian Gaines and Wolfgang Poppelbaum.
- It uses streams of random bits to perform computations, offering advantages like noise immunity and simpler hardware compared to analog computing.
- Stochastic computing was competitive with supercomputers in the 1960s and was used in early neural networks and probabilistic models.
- Modern applications could include energy-efficient deep learning, with potential benefits like 10,000x energy efficiency improvements.
- Extropic, a startup, may be working on stochastic computing, but their communication is unclear.
- Quantum computing companies like D-Wave may effectively be building stochastic machines, though their marketing focuses on quantum woo.
- Stochastic computing faces challenges like bit generation and storage but remains a promising alternative for specialized tasks.