AI (HIERARCHICAL TEMPORAL MEMORY OR HTM)

Hierarchical Temporal Memory is inspired by the capability of the human brain to process ‘high order sequences’ – that is, continuous and concurrent streams of data from multiple sources. Accordingly, this technique shines in detecting not just static patterns, but time-based patterns from data using multiple sources simultaneously. This capability is essential to human intelligence. This capability may also be relevant in designing cognitive analytics on continuous streams of multi-source data generated in IoT-based systems (including Industrial IoT).

In this training, we will use as a reference, open source Numenta HTM, an implementation of HTM built over the last fifteen years (and evolving). We will first present a conceptual overview of HTM, and then run a live experiment for pattern recognition and anomaly detection on sensor data, to help translate the theory into practice. We will also introduce attendees to Numenta HTM Studio that may be suitable for non-technical business analysts.