Inference circuit for improving online learning
AssigneeHRL Laboratories, LLC
InventorsKarl P. Dockendorf, David W. Payton
Life of this patent
Estimating the term…
Abstract
Described is an inference circuit for pattern recognition for use within convolutional neural nets for online learning using K-means clustering. The inference circuit includes a set of templates, each template having a template data memory. The inference circuit also include at least one match element, the match element being operable for receiving an input data pattern and determining a degree of match between the input data pattern and a template. A best match logic circuit is included for selecting a template in the set of templates that best matches the input data pattern, said template being a best match template. Finally, an updated is included for probabilistically modifying the template data memories based on the best match template.
Claims18
An inference circuit for improving learning, the Inference circuit comprising:
- a set of templates, each template having a template data memory;
- at least one match element, the match element being operable for receiving an input data pattern and determining a degree of match between the input data pattern and a template;
- a best match logic circuit, the best match logic circuit operable for selecting a template in the set of templates that best matches the input data pattern, said template being a best match template;
- and
- an updater, the updater operable for randomly modifying each memory element in the template data memory of the best match template by one memory quantization level.
A system for improving learning, the system comprising:
- one or more processors and a memory, the memory having executable instructions encoded thereon, such upon execution of the instructions, the one or more processors performs operations of:
- determining a degree of match between an input data pattern and a template with a set of templates, each template having a template data memory;
- selecting a template in the set of templates that best matches the input data pattern, said template being a best match template;
- and
- randomly modifying each memory element in the template data memory of the best match template by one memory quantization level.
A method for improving learning, the method comprising acts of:
- determining, with a match element, a degree of match between an input data pattern and a template with a set of templates, each template having a template data memory;
- selecting, by a best match logic circuit, a template in the set of templates that best matches the input data pattern, said template being a best match template;
- and
- randomly modifying, by an updater, each memory element in the template data memory of the best match template by one memory quantization level.
Dependencies are inferred from the claim text.