There are 4 main types of machine learning:

Supervised Learning

Humans feed labeled training data into the algorithm to teach the AI how to predict the output label based on new input data. Labeled training data is often used. Typically supervised learning problems are classification or regression problems. Training data sets are very important, as is testing data sets

  • Decision Trees
  • Classification
  • Linear Regression
  • Nearest Neighbor
  • Bayesian Inference

Unsupervised Learning

Algorithm is trained with unlabeled data. There is no teacher and the algorithm discovers relationships, classifications and clusters/groups by itself

  • Self-organizing maps
  • Nearest-neighbor mapping
  • k-means clustering
  • Singular value decomposition

Semi-supervised Learning

A mixture of labeled and unlabeled data is used for training the AI. Methods and algorithms from both supervised and unsupervised data is combined.

Reinforcement Learning

The AI learns by observing the environment and it tries to maximize the reward and minimize the risk. The AI depends on a feedback loop to indicate the reward or risk for every action that it takes. Learn policy

Agent, Environment, Action

  • Variational Autoencoders (VAE) – Good luck finding a comprehensable description of an VAE
    • Encoder – Neural network that takes in data and generates a hidden representation using weights and biases
    • Decoder – Neureal network that takes the hidden representation and generates the data
    • Loss Function –
  • Generative Adverserial Networks (GAN) – 2 Neural networks competing with each other
    • Generator – convolutional network
    • Discriminator – deconvolutional network