Having just got back from CPPCON 2018, I have a new fervor to try something new. I have decided to dive headfirst into Machine Learning and Neural Networks.

My journey starts with watching some videos:

But What is a Neural Network?: 3Blue1Brown

https://youtu.be/aircAruvnKk

This series of videos linked me to

Neural Networks and Deep Learning

http://neuralnetworksanddeeplearning.com/

Which I am currently consuming.

Some Thoughts:

In the late 1990’s I had an idea for an image compression algorithm with this simple idea:

Imagine, for each row of pixels, I could find a random seed that could generate that row.

So I wrote a quick app that would increment a seed and determine a fitness score based on how well the generated pixels matched the actual pixels.

This work “okay” and after hours of computation, I was actually able to get an image that resembled the source. (The picture was a black and white photo of Alfred Hitchcock with a cigar in his mouth and a bird at one end).

I could never get it to work perfectly and I determined I needed a better random number generating function.

Work and life happened and I haven’t returned to that project.

How this relates is how the process of training a Neural Network is very similar to what I was doing 20 years ago. I wish I would have continued along that path back then.