TinyML Talks - The New Neuromorphic Analog Signal Processor Concept and Technology Platform

This talk presents the Neuromorphic Analog Signal Processor concept, explaining how trained neural networks are converted for analog inference on CMOS chips. It discusses the hybrid analog-digital architecture, low-power sensor applications, design flow, and deployment considerations.

Syllabus

Introduction
About tinyML
Core of technology
Physical layout
Design flow
Business model
Product line
Human Activity Recognition
Digital Fingerprint
Narrow Voice
Data Transfer
Partnership
Can you elaborate further on how the train neural network is converted to an AP
Can resistors representing weights be reprogrammed
Analog arrays vs gate arrays
Job opportunities
Typical number of neural network parameters
How is neural sense better than noise cancellation
Can neural net recognize music
Is the digital imprint application agnostic
Is there any correlation to neuromorphic processors
How can we use these models
What kind of tools can we use
Simulations
Direct sensor to spike converters
Publications
Mismatch
More questions
Precision
Conclusion
Strategic Partners