Abhi presents his work over the last year, looking at how dendritic architectures can help with deep learning, and then into how the spatial pooler, temporal memory and grid cells can be applied in the Monty system.
00:00 Introduction 01:13 Dendrites 10:54 Spatial Pooler and Temporal Memory 23:05 Temporal Memory for Monty 28:28 Grid Cells for Monty
Interesting presentation. Two coins to Abhi from my side:
you claim better to use C++ for acceleration than Python and try to port Numenta HTM into C++. Maybe you can take a look at HTM-Community Github, where HTM.core in C++ was much improved, optimized and further developed.
