Keras TF implementation of Adaptive Locally Connected Neuron: FocusingNeuron
Repo for focusing neuron (adaptive locally connected neuron)
Paper: https://arxiv.org/abs/1809.09533 Notes:
I have added keras implementations and some new ipynb for experiments:
Kfocusing: the focusing neuron layer class file, includes tests for simple 2-hidden layer Focused network and CNN+FCS architectures. (Requires included keras_utils.py)
KfocusingTransder: test focusing neuron in transfer learning with keras.applications and pretrained models (VGG1-16)
Boston experiment notebook
Reuters experiment notebook ()
Random_Syntethic_Tests-master-forPaper-ready.ipynb repeats the synthetic experiments
Focusing_Network_Test_Single_Run-Mnist-For-Paper-ready.ipynb repeats a single run MNIST experiment
USE dense-nn-weights-mnist-eng-ready.ipynb to experiment on Dense network Weights with noise PADDED MNIST
NOTE Experiments notebooks can be run in GOOGLE colab
I have added tensorflow 2. folder which includes tf2.keras codes.
This version is working but optimizers are not 100%. I can't set
separate learning rates for MU and Sigma.
In addition, I have weight-share option. The neurons can share their
weights now. sharedWeights parameter in FocusedLayer1D now controls
number of distinct weight sets (if it is 0, nweights=num_inputs x num_units)