In the second session we will take a look at neurons, both the biological original as well as the computational imitation: the perceptron. You will gain insights how decisions byy the perceptron are made and how a connection of multiple artificial neurons can enable non-linear decisions. Together, we will derive the calculation for such a multi-layer perceptron.
Further, we will regard the psychological perspective of neuro-cognition, looking at theories like geons (Hummel & Biederman 1992) and the bigram detector, which tried to explain, how objects and words are perceived. Considering the approaches from both disciplines, we try to approach the questions: How does neural learning work? How close/ far is the computational imitation?
- Read the extract of "The Handbook of Brain Theory and Neural Networks" , which will introduce the biological neuron. The text is quite detailed on the biological side, so the goal is to scan the text. Make sure to read enough of the text to understand the depictions 1 and 2.
- After reading the text, watch the video on "Neurons Are Slow! - Machine Learning Is Not Like Your Brain" .
- Read the MIT Mind article: "Is your brain a computer?"
- Rain-friends data: https://www.desmos.com/calculator/3sknu8jqwa
-
Wind-rain NN data (task):
https://www.desmos.com/calculator/hncih8szkd
- Solution (simple): https://www.desmos.com/calculator/1czmp32hjs
- Solution (complex): https://www.desmos.com/calculator/4dtlyi0bjg