Feedback mechanisms in sequence learning systems with temporal processing capability
AssigneeNumenta, Inc.
InventorsJeffrey C. Hawkins, Subutai Ahmad
Life of this patent
Estimating the term…
Abstract
Embodiments relate to a first processing node that processes an input data having a temporal sequence of spatial patterns by retaining a higher-level context of the temporal sequence. The first processing node performs temporal processing based at least on feedback inputs received from a second processing node. The first processing node determines whether learned temporal sequences are included in the input data based on sequence inputs transmitted within the same level of a hierarchy of processing nodes and the feedback inputs received from an upper level of the hierarchy of processing nodes.
Claims20
A computer-implemented method comprising:
- obtaining, by a first processing node, a series of sparse distributed vectors over time based on input data received at the first processing node that changes over time, wherein the series of sparse distributed vectors include non-binary scalar values;
- performing temporal processing on the series of sparse distributed vectors at the first processing node by at least:
- generating vectors indicating activation of a select subset of elements of the first processing node, the vectors generated responsive to receiving the series of sparse distributed vectors by predicting or detecting one or more spatial patterns or one or more temporal sequences in the series of sparse distributed vectors, and
- providing first sequence inputs to another subset of elements in the first processing node or a processing node other than the first processing node, the first sequence inputs representing the activation of the select subset of elements in the first processing node;
- generating output data from the generated vectors to indicate at least prediction or detection of the one or more spatial patterns or the one or more temporal sequences in the input data.
A non-transitory computer readable medium comprising:
- a processor;
- and
- a memory operably coupled to the processor, the memory storing an instance of a first processing node, the first processing node configured to:
- obtain a series of sparse distributed vectors over time based on input data received at the first processing node that changes over time, wherein the series of sparse distributed vectors include non-binary scalar values;
- perform temporal processing on the series of sparse distributed vectors by at least:
- generate vectors indicating activation of a select subset of elements of the first processing node, the vectors generated responsive to the select subset of elements in the first processing node receiving the series of sparse distributed vectors by predicting or detecting one or more spatial patterns or one or more temporal sequences in the series of sparse distributed vectors, and
- provide first sequence inputs to another subset of elements in the first processing node or a processing node other than the first processing node, the first sequence inputs representing the activation of the select subset of elements in the first processing node,
- generate output data from the generated vectors to indicate at least prediction or detection of the one or more spatial patterns or the one or more temporal sequences in the input data.
Dependencies are inferred from the claim text.