# Neural networks

**Lecturer**: Nikolas Bernaola

**Date**: 10/04/2018

**Time**: 17:00

**Place**: Room 209 (2nd floor, Faculty of Mathematics UCM)

**Abstract**:

Machine Learning is the area of study concerned with the statistical analysis of data in order to find patterns and do predictions. Neural Networks are one of the most prominent methods. Inspired on the way biological neurons compute their activations from a weighted combination of inputs, we build a whole array of these neurons to do prediction tasks on some training data and use a weight correction technique to make the neural network approximate the underlying distribution of the data.

In this talk I will present a wholly functional neural network for a classification task, showing the building blocks in Keras. I will explain how the network functions and how the weights are adjusted. After this, I will talk about Deep Learning and Convolutional Neural Networks as an improvement over the original framework. And finally, if time allows, I will briefly disccus the theory behind the effectiveness of these methods.

## Bibliography

- Ian Goodfellow
*et al*.*Deep Learning*. MIT press. Online - Andrew Ng.
*Machine Learning on Coursera*. Stanford University. Course webpage - Andrew Ng.
*Deep Learning Specialization*. Coursera. Webpage