This repo contains an implementation of Hierarchical Temporal Memory in OpenCL.
TODO List:
- Adapt KMA for use with PyOpenCL and test it
- Test it (something near 6 millions malloc calls per second)
- Fix that bug with code failing on Nvidia GPU. It was caused by wrong objects alignment in memory in case of 64-bit GPU.
- Fix that error with being unable to allocate more than 128 megabytes on x64 GPUs (probably fixed).
- Implement random
- Test it
- Implement generic list
- Test it
- Implement input layer
- Implement spatial pooler
- Implement temporal pooler
- Implement classifier
- Implement multi-layer architecture
- Write a proper documentation