Elements of Reinforcement Learning

The five classic elements of reinforcement learning are Policy, Value, Reward and Agent.Additionally, there are two basic blocks that describe the environment and how the agent can influence it. States and actions.A state describes the world to our agent and actions are a predefined set of responses that an agent can choose from. Task The task… Continue reading Elements of Reinforcement Learning

Gaussian Mixture Models

Gaussian Mixture Models assume that the data we are clustering comes from a mixture of Gaussian Distributions (also called Normal Distributions). If you know k-means, you can think of Mixture Models extending k-means by accounting for the clusters’ shape and density. (Remember, k-means was only able to deal with spherical clusters and density, we did… Continue reading Gaussian Mixture Models

Fighting Vanishing and Exploding Gradients

If you repeatedly multiply small numbers with one another they eventually vanish.Similarly, if you keep multiplying numbers > 1 they will eventually explode. In Backpropagation, we multiply partial derivatives with one another over and over again. And this is why the more hidden layers a network gets, the more often we will encounter vanishing and… Continue reading Fighting Vanishing and Exploding Gradients

The Perceptron

Imagine we want to classify Apples as sweet and sour. For some reason, we can do so by using their radius and redness. is the plane we wish to find to separate the Apples.If we classify as negative (sour) and if we classify as positive (sweet). Here are the weights, and is a vector containing… Continue reading The Perceptron

Neural Network Basics

We’ve been through logistic regression here. A Neural Network is basically several logistic regressions stacked on top of each other.The input of each neuron (or perceptron) is the output of some previous neurons. Activation Functions Activation functions are crucial. They help us create nonlinear decision boundaries, add meaning to our features and offer relief from… Continue reading Neural Network Basics