the samples are described by a vector of features

we need to split the data into a training set and a test set

the data in the training set must be labeled

we need to evaluate the results using various metrics

visual recognition problem

unsupervised learning problem

classification problem

association problem

supervised learning problem

The training set must be representative of the real world.

The selected features must accurately describe the sample.

The more features describe the sample the better the model.

a robot simulating the natural behaviour of a rat

an early implementation of perceptron

a maze-solving device functioning on feedback loops

has low accuracy on training set

is underfit

is overfit

is sensitive to small fluctuations of features

fails to find the regression line if the dataset doesn't meet specific requirements

always finds the regression line which minimizes the sum of squares of differences between the actual value and the predicted value

always finds the regression line which minimizes the sum of absolute values of differences between the actual value and the predicted value

requires that the data should follow the normal distribution

always finds the regression line which represents best the actual trend in the dataset

actions of the agent have no impact on the environment

the aim of the agent is to learn all possible states of the environment

agent interacts with the environment

the aim of the agent is to maximize the reward

Logistic Regression

K-Nearest Neighbors

Linear Regression

K-Means