Anomaly detection – problem motivation Anomaly detection is a reasonably commonly used type of machine learning application Can be thought of as a solution to an unsupervised learning problem But, has aspects of supervised learning What is anomaly detection? Imagine you’re an aircraft engine manufacturer …

Motivation 1: Data compression Start talking about a second type of unsupervised learning problem – dimensionality reduction Why should we look at dimensionality reduction? Compression Speeds up algorithms Reduces space used by data for them What is dimensionality reduction? So you’ve collected many …

Unsupervised learning – introduction Talk about clustering Learning from unlabeled data Unsupervised learning Useful to contras with supervised learning Compare and contrast Supervised learning Given a set of labels, fit a hypothesis to it Unsupervised learning Try and determining structure in the data …

Support Vector Machine (SVM) – Optimization objective So far, we’ve seen a range of different algorithms With supervised learning algorithms – performance is pretty similar What matters more often is; The amount of training data Skill of applying algorithms One …

Machine learning systems design In this section we’ll touch on how to put together a system Previous sections have looked at a wide range of different issues in significant focus This section is less mathematical, but material will be very …

Deciding what to try next We now know many techniques But, there is a big difference between someone who knows an algorithm vs. someone less familiar and doesn’t understand how to apply them Make sure you know how to chose the best …

Neural network cost function NNs – one of the most powerful learning algorithms Is a learning algorithm for fitting the derived parameters given a training set   Let’s have a first look at a neural network cost function Focus on application …

Neural networks – Overview and summaryWhy do we need neural networks? Say we have a complex supervised learning classification problem Can use logistic regression with many polynomial terms Works well when you have 1-2 features If you have 100 features …

The problem of overfitting So far we’ve seen a few algorithms – work well for many applications, but can suffer from the problem of overfitting What is overfitting? What is regularization and how does it help Overfitting with linear regression …