conx is a small neural network library at https://github.com/ArtificialIntelligenceToolkit/conx that stopped working when keras got absorbed into tensorflow.
So I cloned it and fixed it. But you will need a python 3.7 based distribution.
If you have an Apple "M" chip, this may be a problem because you will not be able to use a python 3.7 distribution with tensorflow. I plan to fix this problem in the future. For now, please run on Windows or Mac with Intel chips and use python 3.7-based environment.
All you need to do is to create a new python 3.7-based environment called conx2 (or anything else you wish), either through anaconda navigator or in an anaconda shell with conda create -n conx2 python=3.7.
Then, open an anaconda shell and activate that environment with conda activate conx2, followed by pip-installing conx2 from github like so:
python -m pip install git+https://github.com/dinorows/conx2
Then, follow by installing tensorflow:
pip install tensorflow
You probably also want to:
pip install pyqt5
In order to plot conx loss and accuracy graphs straight to your notebook while your network is training, add a cell with this code in your jupyter notebook:
import matplotlib
import matplotlib.pyplot as plt
matplotlib.use('Qt5Agg')
%matplotlib inline
Make sure to run your notebook within your new environment by switching to your conx2 kernel.
Other than that, your old conx-based python notebooks should not need to be modified at all. In other words, import conx, not conx2:
import conx as cx
and keep on computing!