On the fourth day Hasan showed us the ‘magic behind the curtain’ in regards to the processes behind Machine Learning. He described machine learning as a way to get things done, from input to output but the way in which they do this is the magic. He then took us through the very basic workings of machine learning model training. Using his love for sugar, in particular Mcdonalds ice cream as an example.
‘Once the machine finds the right fit it can predict the future’ Hasan S
y = wx + b
We learnt that this simple equation above is the fundamental maths behind machine learning, and disussed two types of ML models – regression and classification. In day 2 of the fellowship we used a classification model to create our interactive poems, training the model on webcam input. Hasan then went on to explain the importance of the ‘rate of learning’ , if you choose a model that is too fast the machine will miss the correct solution, if it’s too slow it will take a very long time to reach the solution.
‘Machines are not limited to the 3D world they can go into the billions of dimensions/imputs’ Hasan S
Hasan expained the workings of artifical neural networks and the fact they are basically a mathematical formula. A layering of these ‘neurons’ ontop of each other is what we call ‘deep neural networks’. The machine creates a filter where the data is passed through at each layer in the NN, crunching the numbers down. Artifical NN are roughly inspired by the workings of the human brain however thier design is based on a singal biological neuron. Another analogy for the processes behind ML is a ‘Universal Function Aproximator’ , we discussed that there may need to be a better analogy for it still.
‘How do you know a dog is a dog, what is the formula for that?’ Hasan S
We discussed the differences and similarities between human learning and machine learning, touching on reinforced learning where the machine repeats the machine learns to achieve a goal in an uncertain, potentially complex environment employing trial and error to come up with a solution to the problem.
ORDER / CHAOS SIGNAL / NOISE
After the break Hasan took us through an intro to p5.js. We were then asked to think of our reference to order/chaos signal/noise we had been asked to find for this task. My example was the ‘edge of chaos’ a transitional space between order and disorder that is hypothesised to exist in a wide variety of systems. It is a metaphor in sciences that some physical, biological, economic and social systems operate in a region between order and complete randomness/chaos, where complexity is maximal. I found this a really interesting concept when thinking about our previous discussions around human thinking and/or consciousness, and is something I wish to explore further in my own artistic research.
This was ofcourse hard to represent in a basic p5.js sketch, but I created some nice visual experiments using line, colour, randomness, and mouseClicked event triggers changing the background colour and placing circiles.