Logistic regression is a classification technique for classifying data points x (a vector) into one of K classes . It works by (essentially) projecting the datapoints onto a set of (pre-specified) features (which are simply vectors formed of functions of the datapoints’ components), and then finding linear separating (hyper-)planes in…

# Classification: Introduction

In this series of four (and eventually possibly more) blogs, I am going to look at classification methods, and in particular (at least in the first instance) I am going to look at neural network-type methods. This is a hot topic (again) at the moment, with the recent demonstration of…

# Intentionality Prediction

I’ve recently finished and submitted a paper on the interesting topic of intentionality prediction, that is, figuring out the intent of something from the action that it is taking. In the paper (one of a number that I have published in this area), we look at how we can infer…

# Hadoop Part IV: Online Parameter Estimation

For the final (for now) part of this series, I am going to extend the particle filter to do online parameter estimation using online Expectation-Maximization (EM) to calculate an estimate of the autoregression parameter at each stage of the particle filter. There are many options for online parameter estimation, including…

# Hadoop Part III: Multiple Output in Hadoop

The output of the Hadoop MapReduce particle filter from the previous post is simply a list of doubles giving the state for each particle after resampling. This is not ideal because this post-resampling particle collection is a more crude representation of the post-observation state posterior than the pre-resampling, weighted collection. Obviously…

# Hadoop Part II: Particle Filters in Hadoop MapReduce

In this article, I’m going to go about implementing a basic particle filter in Hadoop MapReduce. This is really just a personal interest project for me to get started learning Hadoop based on an algorithm that I am familiar with and suits MapReduce (to some extent), but this might have…

# Hadoop Part I: Configuring Hadoop (for Complete Beginners)

When starting out with new technology, I often find that one of the most challenging bits is getting things into a place where I can start to write code. I usually find this process quite frustrating and it isn’t a part that I particularly enjoy. Now, this may say something…

# Hadoop Introduction: Using Hadoop for Sequential Monte Carlo (Particle Filter) Parameter Estimation

Recently, I have become interested in parameter estimation for sequential Monte Carlo (SMC) methods (primarily, but not exclusively, particle filters), due to writing about them for my forthcoming book (obligatory plug – sorry). And I’ve also become interested in learning about Hadoop (an interest not uncorrelated to my current job…

# Google Cardboard

I’m very excited to have just built a working Google cardboard. I’ve been looking forward to VR since it first became visible in the mid(?) nineties. And now that I’ve tried it for the first time, I can say that I am more excited about it than ever. It is…

# A Brief Review of “Collaborative Filtering for Implicit Feedback Datasets” by Y. Hu, Y. Koren and C. Volinsky

During a recruitment process that I took part in, I was asked to write a review of the paper Collaborative Filtering for Implicit Feedback Datasets (2009) by Y. Hu, Y. Koren and C. Volinsky. In the review I gave a Bayesian interpretation of the model, which makes some of the parameters…