Saturday, September 17, 2016

Neural Networks and all of that


More than one young person  ( I mean under the age of 20) recently solicited  Marcus and Agata's views on Further education.  We were flattered. I would say that Marcus never regretted pursuing a Science/ Engineering based study career and here are some of the reasons why:

- It was actually going to be useful after graduation.
- We never bought that shit luxury argument that you should goto University to study anything on a whim (please take note UK students who made Make Up the most popular course in 2016.  I am not kidding).  Because you are being told how to think, and reason whatever the course.  Hmmm.
- Also a subject where Logic and Calculation and Research and Thinking and Invention are involved
- When you know the fundamentals, e.g. Rules of Mathematics, Logic, Computer Design, you can build on that to predict outcomes of your designs.   I'm not learning a subject which is 'merely' backed by observational data to say 'this is the way the world appears to be'
- Oh and Software Simulation,  The tool that allows every Engineer to prototype and not blow things up in the physical world, a life-saver


Back in the Day
Actually Marcus has had an interest in what was called AI (Artificial Intelligence) for well,  some time,  in fact since the term coined in the 1980's.   My Masters' thesis was indeed to develop a  rule based tool that could be used to detect the correct actions when powerline breakers overloaded and switched.  It was coded in both Prolog and Lisp, the final code being mapped to pure Prolog.

In 2016 with the enormous increase in compute power the Machine learning game has rather moved on.

In the 1980 you would quiz experts and via question and answer sessions tease out their hidden secret rules and code them, into Expert Systems, or systems that appear to exhibit intelligence.  Today the shift is to look at the vast quantities of data, and devise general techniques to use the raw data itself, to find answers, or rules of behaviour.

By watching just 3 videos below I hope you will be enlightened and educated.   It might even charm you towards studying Philosophy, Logic, Mathematics, Computer Science, Programming and Design, Manufacturing.  

 
3 video tutorials


01. Brais Martinez: Deep Learning





02.Mike Pound: Blurs and Filters with Kernel Convolution




03. Convolutional Neural Networks






Bonus Video
K-Means