HierarchicalTemporalMemory.jl
HierarchicalTemporalMemory.jl copied to clipboard
A simple, high-level Julia implementation of Numenta HTM algorithms
https://www.frontiersin.org/articles/10.3389/fncom.2017.00111/full
A tester unfamiliar with HTM should try to use this package and note where the docs are incomprehensible, lack critical info, or the interface doesn't make sense. Tester workflow: 1....
https://juliaquant.github.io/MarketData.jl/stable/company_financial_series.html
https://juliastats.org/TimeSeries.jl/stable/getting_started/
https://github.com/Oblynx/htm.jl/blob/421036ce9ee16eb4c6241b8ccf788a4f624daa95/src/dynamical_systems.jl#L240 .rowval is used, which is dangerous as the array slice could potentially be a SparseVector and not a SparseMatrixCSC
Adding single new segments to TM.DistalSynapses is very slow, so maybe cache many segment additions and perform them together.
Performance expectations: - `sp_activation` - Dense: no effect - `adapt!` - Dense: strided memory access, might degrade performance