Neural network built from scratch with python and numpy
Two great resources to get you started with machine learning are:
- Andrew Trask’s “Grokking Deep Learning” I am Trask - a book being used by the Machine Learning Foundations course at Udacity.
- Sira Raval’s youTube channel - fast, funny, inspiring and used for the basis of the Udacity Mooc’s course Machine Learning Foundations.
There are a lot of libraries available (like Tensorflow) to help build Neural Networks, however it is important to understand the basic building blocks. This is especially important for machine learning which requires various aspects of statisitics, mathematics, computing, information science and some understanding of the hardware available (STEM):
- Data manipulation - an understand of how to shape the data
- Hyper parameters - how to tune the network to achieve convergence in a suitable time with the avialable computing power
- Gradient descent - to converge towards the optimal solution
- Back propogation - to train the network and update it’s weights
The data comes from the UCI Machine Learning Database.
