BAMI: HTM Overview

HTM is inspired by neocortical cells but only has an interest in emulating parts needed for intelligence.

HTM uses bitwise operations.

Where ANNs rely on summing real numbers to decide if they will fire an HTM uses a bitwise OR on incoming context.

The neocortex is largely similar while certain regions have slightly adjusted sizes of dendrites or counts of certain cell types.

"Encoders" are used to convert sensory data in to an SDR for the system to use.

Online learning: HTMs learn while the system is online and in use; it looks at the predicted world state and backpropagates (via hebbian learning) based on what the actual next world state was.