This page contains technical definitions of few very common terms in Machine Learning, Reinforcement Learning and Optimization in general. Definitions are in random order.

1. Inverse Reinforcement Learning. [Russell 1998]: Determine the reward function that an agent is optimizing, given, a) Measurement of the agent’s behavior over time, in variety of circumstances b) Measurement of sensory input that agent; c) a model of the environment.

2. Generative Model: Given observed data and target generative model models both the input and output. Meaning, for finding, P(Y|X) a generative model will find P(X|Y) and P(Y) from the data and then multiply them to get P(Y|X)  according to Bayes rule. Therefore this explains simply what’s written on wiki,  a generative model is a model for generating all values for a phenomenon, both those that can be observed in the world and “target” variables that can only be computed from those observed.

3. Discriminative Model: Unlike generative model, discriminative model models directly P(Y|X) from data without computing P(X|Y) and P(Y).  Therefore, (According to wiki) By contrast, discriminative models provide a model only for the target variable(s), generating them by analyzing the observed variables. In simple terms, discriminative models infer outputs based on inputs,