NeuroMem CM1K chip
The CM1K is a neuromorphic chip opening new frontiers for smart sensors and cognitive computing applications. It can solve pattern recognition problems from text and data analytics, vision, audition, and multi-sensory fusion with orders of magnitude less energy and complexity than modern microprocessors.
The CM1K chip features 1024 interconnected neurons working in parallel and capable of learning and recognizing patterns in a few microseconds. The neurons behave collectively as a K-Nearest Neighbor classifier or a Radial Basis Function. They are trainable and especially suitable to cope with ill-defined and fuzzy data, high variability of context and even novelty detection. Last, but not least, multiple CM1K chips can be daisy-chained to scale a network from thousands to millions of neurons with the same simplicity of operation as a single chip. The CM1K is a significant evolution of the Zero Instruction Set Computer invented in 1993 by IBM and Guy Paillet, CEO of General Vision. Download datasheet.
CM1K was awarded the New Product Innovation Award for Cognitive Computing Processors by Frost and Sullivan.
- All neurons have the same behavior. They receive and execute the same instructions in parallel.
- All neurons are interconnected. The “winner-takes-all” and novelty (in the absence of winner) can trigger autonomous learning. Confusion can be corrected.
- The neurons can behave as a K-Nearest Neighbor (KNN) classifier or a Radial Basis Function (RBF).
- Patterns are broadcasted to all neurons at once, whether for learning or recognition.
- Recognition and learning latencies are constant, in the order of micro-seconds per pattern and independent of the number of neurons in use.
- Neurons are trained by example and decide autonomously when it is necessary to commit a new neurons and/or correct their own firing threshold (influence field)
- The knowledge built autonomously by the neurons can be saved and restored