Machine Learning 101

A short talk introducing the field of machine learning touching on supervised learning, unsupervised learning and reinforcement learning.

Transcript

1. Machine Learning 101 How do machines learn?
2. Computational Neuroscience Artificial Intelligence ML
3. Supervised Learning Unsupervised Learning Reinforcement Learning
4. Supervised Learning { , ,… , ( , )} :   → Data: Model: : : Vectorised images, videos, audio Labels of respective data points Good models generalize to unseen data Model fits Vectors to Labels by finding decision boundaries Classification
5. Supervised Learning Models: Regression, Bayesian, Decision Trees, SVMs, Artificial Neural Networks 1) Large data 2) Better hardware Classification
6. Unsupervised Learning {   , … , ( )} Data: Model: : Vectorised images, videos, audio No Labels! Captures distribution of data, Compresses data, Classify Data Mysterious: Describe hidden structure
7. Unsupervised Learning Clustering Mysterious: Describe hidden structure Generative Adversarial Network
8. Reinforcement Learning Agent: Take Actions in Environment to maximize cumulative award. Policy used to determine action in a state Policy learned from received reward Exploring Environment maximizing reward State
9. Reinforcement Learning Playing Othello: Exploring Environment maximizing reward State 64x1 1 black 0 empty -1 white
10. What can ML do for CN? Perhaps ML will teach us more about the brain? Who knows! Convolutional Neural Network
11. There is hope Mandelbrot set Simple formula: 4 = + Perhaps the brain’s endless intricacy can be modelled by simple rules.