Pattern recognition and perceptrons, an interesting lesson - BASIC Hacking - 13

In this video, I introduce the problem of pattern recognition performed using a perceptron. The concept of perceptron is first explained and, then, the problem of pattern recognition is described. Two solutions for this problem are discussed: one which is based on a numerical training method, and the other which is analytical and utilizes exact formulas, i.e. no number crunching is required. This represents an interesting lesson in machine learning: sometimes it is possible to find and utilize analytical solutions, although this might not seem the case initially. The exact approach is, eventually, implemented in BASIC on the ATARI 800XL. Finally, to clearly see how it works in practice, several examples are presented.
Please, don't forget to subscribe, comment and share with others, if you like it. This will definitely encourage me to create more videos like this!
Thanks!
Useful links:
www.virtualdub...
atariwiki.org/...
github.com/lib...
www.gnu.org/so...
Previous video:
• Autoencoders in ATARI ...
Background music:
Eric Skiff, Underclocked. Available at EricSkiff.com/m...
#BASIC #programming #hacks #8bit #Commodore #Atari #Amiga #MSX #machinelearning #quantumcomputing #numericalmethods #algorithm #retrocomputing #physics #mathematics

[@mad_circuits]

Here we go! That is what I'm talking about! Good video, well done! Maybe we can increase complexity to recognize numbers by OCR. 5x7 char bitmap? What do you think?

[@jeanmichelsellier]

That would be a great idea for a next video but it will take some time to prepare it. Thanks a bunch for sharing your thoughts! This is really appreciated and helpful!