Abstract: An adaptive pattern recognition system optimizes an invariance objective and an input fidelity objective to accurately recognize input patterns in the presence of arbitrary input transformations. A fixed state or value of a feature output can nonlinearly reconstruct or generate multiple spatially distant input patterns and respond similarly to multiple spatially distant input patterns, while preserving the ability to efficiently evaluate the input fidelity objective. Exemplary networks, including a novel factorization of a third-order Boltzmann machine, exhibit multilayered, unsupervised learning of arbitrary transformations, and learn rich, complex features even in the absence of labeled data. These features are then used to classify unknown input patterns, to perform dimensionality reduction or compression.

Abstract: An HTM node learns a plurality of groups of sensed input patterns over time based on the frequency of temporal adjacency of the input patterns. An HTM node receives a new sensed input, the HTM node assigns probabilities as to the likelihood that the new sensed input matches each of the plurality of learned groups. The HTM node then combines this probability distribution (may be normalized) with previous state information to assign probabilities as to the likelihood that the new sensed input is part of each of the learned groups of the HTM node. Then, as described above, the distribution over the set of groups learned by the HTM node is passed to a higher level node. This process is repeated at higher level nodes to infer a cause of the newly sensed input.

Abstract: An HTM node learns a plurality of groups of sensed input patterns over time based on the frequency of temporal adjacency of the input patterns. An HTM node receives a new sensed input, the HTM node assigns probabilities as to the likelihood that the new sensed input matches each of the plurality of learned groups. The HTM node then combines this probability distribution (may be normalized) with previous state information to assign probabilities as to the likelihood that the new sensed input is part of each of the learned groups of the HTM node. Then, as described above, the distribution over the set of groups learned by the HTM node is passed to a higher level node. This process is repeated at higher level nodes to infer a cause of the newly sensed input.