MachineLearning VS HMake

Softmax is at the end of this source file: https://github.com/aromanro/MachineLearning/blob/master/MachineLearning/MachineLearning/ActivationFunctions.h

Continuing extending the machine learning project: https://github.com/aromanro/MachineLearning With dropout I managed to go over 99% accuracy on digits from the EMNIST dataset, currently I'm thinking of adding batch normalization and then adding convolutional networks.

I have a project on GitHub on this: https://github.com/aromanro/MachineLearning

A project going up from simple linear regression to neural networks (multilayer perceptron for now, applied on xor, the iris dataset and EMNIST dataset): https://github.com/aromanro/MachineLearning

I implemented a multilayer perceptron in C++ (using Eigen for matrix computations): https://github.com/aromanro/MachineLearning

Fork from here, then first build Example 1 from here as per instructions. Then create build-dir in SFML and run hhelper twice and hbuild once there.

I wanted to include a modules Hello World example in my project https://github.com/HassanSajjad-302/HMake but I could not. I considered 2 options. 1) Using the STL module source, but that was not available. If it had been available, only one line would have been needed. ADD_MODULE_DIR_TO_TARGET(app, "./modules/", ".*");

I am writing a new build system HMake on top of CMake. Basically, HMake is a C++ library that generates a CMakeLists.txt file. There are very important features missing in it currently. Though I believe they can be added. I was able to generate two trivial CMakeLists.txt files which work for my two projects. One example is the following.