Research Glossary

Here we give definitions of the technical terms used in the research wiki. Some terms are defined in the glossary section of the HTM White Paper

Bursting - A situation in which no cells of an active column predicted the column’s activity. In this case, all of the columns’ cells are set to active. This signifies that all temporal contexts are possible. Subsequent activity will eliminate the incorrect contexts and lock in to the sequence that is occurring, if any.

Dendrite segment - See the HTM White Paper glossary.

Flash inference - Humans can identify objects in images that are flashed very briefly in front of them. Because there is not enough time for a saccade, we know that the brain, and thus our algorithms, should be able to activate stable representations at higher levels with a single "flash inference."

Hysteresis - Hysteresis in the algorithms is where activity is persisted. This occurs in different contexts but may be a necessary functional property of neurons in certain areas of the cortex.

Inference - ?? Temporal prediction of future cortical region states, specifically the active columns.

Layer - A cortical layer consisting of a set of columns and cells and the algorithms operating on them. This is different from the levels in a hierarchy; each level can contain multiple layers (layers 1-6 of the neocortex).

Mini-burst - A stronger activation of a cell that occurs when it becomes active after having been predicted from the top-down

Sensory-motor inference algorithm -

Temporal pooling - See the HTM White Paper glossary

Temporal memory -