Add in Classical/Standard Competitive (Hebbian or Energy-based) Learning Synaptic Components

[ago109]

Neuroscience components should offer a parallel sub-package to components.synapses.hebbian named components.synapses.competitive where classical competitive Hebbian learning models such as:

• Self-organizing map (SOM) - this is done, actually (and featured in ngc-learn main branch alongside ngc-museum main branch); this still needs to be reformulated a bit further to offer a mini-batch update (currently it's just online)
  ** Target model component = SOMSynapse
• Hopfield network (at least modern Hopfield energy memory model)
  ** Target model component = HopfieldSynapse
• Vector quantization
  ** Target model component(s) = VectorQuantizeSynapse
• Adaptive resonance theory (at least ART-1 and ART-2a; binary and continuous flavors)
  ** Target model component = AdaptiveResonanceSynapse
• A nearest neighbors (iterative KNN) probe (Added to ngclearn.utils.analysis)