CogniMem News
Come see our CogniBlox™ at embedded world 2012 in Nuremburg Germany at the Everspin stand #428 in hall 1, Feb 28th – Mar 1st
See article describing superiority of CogniMem hardware based RBF classifier in performance and accuracy over S/W alone based approaches for Face Recognition!
“A Hardware/Software Co-design Model for Face Recognition using Cognimem Neural Network chip”
By Santu Sardar, Gayrav Tewari,
K.A. Babu
DRDO; Hyderabad, India
See recently published book on CogniMem CM1K (1024 neurons) predecessor ZISC (Zero Instruction Set Computer) ZISC036 (36 neurons) for target tracking!
“Target Tracking with the Zero Instruction Set Computer”
Evaluation of the Computational Capacity of the ZISC in Target Tracking
By Wendall C. Deck
CogniMem releases Volume 1, Issue 2 of the CogniMem Communique'.
CogniMem CM1K Chip
The CogniMem CM1K Chip is the first ASIC version of the CogniMem neural network product line optimized for Cognitive Computing. It features 1024 neurons working in parallel to implement two reknown non-linear classifiers. It can recognize patterns at high speed while coping with ill-defined data, unknown events and changes of contexts and working conditions.
- Trainable: automatic model generator
- Reliable: re-known pattern classifier
- Ease of Use: software tools and SDK
- High Speed: board and chipsets
- Low Power: board and chipsets
- Object Recognition
- Target Tracking
- Signal Classification
- Data Classification, Clustering
- Anomaly / Novelty Detection
- Template Matching
Massively Parallel Hardware Acceleration for Pattern Recognition
When pattern recognition is the problem – CogniMem’s Cognitive Computing chips are the answer. CogniMem is leading the industry with Cognitive Sensing & Computing - going beyond Von Neumann computer architecture to deliver unmatched levels of processing at far less energy.
Intelligence for Sensing
CogniMem enables trainable adaptive filtering for vision, speech and other sensing devices while reducing cost and energy.
Intelligence for Data Mining
CogniMem enables a flexible and scalable architecture natively implementing well-known non-linear classifiers Radial Basis Functions and K-Nearest Neighbor. Latency of the recognition stays fixed independent of the amount of memories being searched at very low power.
CogniBlox
Cognimem CogniBlox™ Reconfigurable module featuring 4 CM1K chips, one Lattice FPGA and 4MB of Everspin MRAM. Stackable and scales to large arrays of contiguous neurons with a constant latency of only 10µs.