Neural Architectures TRIZ Cases
- Neural Network Layer Data Processing via Feature Map Segmentation
A data processing system generates part-input feature maps from input feature maps and convolves these segments with filters to produce output maps.
- Non-linear Data Masking for Secure ML Training
Autoencoder-based non-linear masking transforms sensitive data into encoded forms for machine learning training.
- Reconfigurable Neuromorphic Processor Interconnection Scheme
A reconfigurable neuromorphic processor uses dynamic instruction sets and segmented synapse arrays to execute complex tasks.