A scratch written, regressive/progressive propagation therein gradient descent aligned artificial neural network framework program of mine, that reasonably models human neuronal phenomena.
https://github.com/JordanMicahBennett/NEURAL_NETWORK_PRACTICE/
The source sample includes demonstrations, configured via ( value * value * value ) neuron topology (this is extensible see ...USES); such that such:
i.Perceives handwritten numerals ( 1024x1024x10 neurons ). Conclusively, said neural network aptly perceives priorly unseen input numerals. (including terribly or incompletely written input numerals)
ii.Perceives xor input vectors ( 2x2x1 neurons ).
This neural network utilizes a simplistic topological mechanism, therein encompassing the construction of sequentially hierarchically horizontal patterns of knowledge genera/perception.
See INSTRUCTIONS.md.