Column/assembly graph neural network

Research and Theory
1

I have created a knowledgebase/graph neural network architecture (GIAANN prototype) which can be used in conjunction with a predictive network (eg transformer) to predict the next token in a sequence. It trains a set of columns for every new noun encountered in a textual corpus, and feature neurons for every contextual word (non-noun) directly surrounding each noun. It supports distinctions in dendrite proximity (SANI; sequentially activated neuronal inputs), and is based on neural assembly and cortical column theory.

train visualisation

predict visualisation

3 Likes
2

The “predictive network” component has been removed as top-1 neuron selection based on activation level across the column/network now provides high prediction accuracy.

It is currently citing “Hawkins, J. et al. (2011). Hierarchical Temporal Memory (HTM) Whitepaper (Version 0.2.1). Numenta”, although they may have more recent preferred citations.

2 Likes
3

Hi @richardbaxter, this might be better suited for the HTM forum - https://discourse.numenta.org/

1 Like