HierarchicalTemporalMemory.jl
HierarchicalTemporalMemory.jl copied to clipboard
Oblynx
A simple, high-level Julia implementation of Numenta HTM algorithms
This pull request changes the compat entry for the `Chain` package from `^0.4` to `^0.4, 0.5`. This keeps the compat entries for earlier versions. Note: I have not tested your...
![github-actions[bot] avatar](https://avatars.githubusercontent.com/in/15368?v=4 "github-actions[bot] avatar"){kind=avatar}
This pull request changes the compat entry for the `Setfield` package from `^0.8` to `^0.8, 1`. This keeps the compat entries for earlier versions. Note: I have not tested your...
Regions are a more intuitive concept to use HTM, exemplified in the test [test_projection.jl](test/test_projection.jl). Needs docs.
The SP tiebreaker is necessary to enforce strict sparsity control, but it has undesirable sideffects in other properties: - `sp(z)` becomes stochastic (random seed from hash(z)?) - noise robustness suffers...
Numenta's latest paper [Grid Cell Path Integration For Movement-Based Visual Object Recognition](https://arxiv.org/abs/2102.09076) implements an encoder for MNIST. In the paper they use a CNN with a k-winner layer. I wonder...
relevant discussions: - https://github.com/alan-turing-institute/MLJ.jl/issues/60#issuecomment-572517203 - [MLJTime](https://github.com/alan-turing-institute/MLJTime.jl) & https://github.com/alan-turing-institute/MLJ.jl/issues/303 - https://github.com/alan-turing-institute/MLJ.jl/issues/51 Some of these things need to be developed on MLJ's side first.
https://github.com/Jutho/SparseArrayKit.jl ## N-D tensors Synapses are by nature 2D (pre-, post-synaptic addresses). In case of multiple spatial dimensions, no reason to burden the synapse matrices with those but rather should...
https://github.com/invenia/NamedDims.jl - Synaptic matrices have presynaptic neurons and postsynaptic dendritic segments: `{pre_n, post_s}`
Local inhibition is currently a costly nonlinear top-k convolution. Maybe the problem can be decoupled into a quick linear filter and a global top-k like @freemin7 's suggestion. The problem...
Data pre-processing (e.g. [here](https://github.com/Oblynx/HierarchicalTemporalMemory.jl/blob/master/test/temporal_memory_test.jl)) is left largely to the user, and may be troublesome or beginner-unfriendly in (potential) neuroscience-related applications. * Automating the process could help? * Integrating a data...
