Surprisingly, quite a few data scientists overlook the importance of linear regression and the problem of overparametrization. In this post, I'm going to describe mathematically what it means for a model to be overparametrized and the general strategies used by a statistician to resolve non-uniqueness.

Ranks, Subspaces, and Bases

Suppose …

Wikipedia defines

The zero-inflated Poisson model concerns a random event containing excess zero-count data in unit time. For instance, the number of insurance claims within a population for a certain type of risk would be zero-inflated by those people who have not taken out insurance against the risk and thus …

Foursquare uses its unique location technology and foot traffic panel to produce personalized recommendations of spatial items such as restaurants. Recently I've read a paper LCARS: A Spatial Item Recommender System that I implemented from scratch in R.

The recommender system combines the querying user's interest and the local preference …

Introduction

The field of spatial statistics is experiencing a period of rapid growth, though less so compared to that of machine learning and artificial intelligence, with the onset of location technology and efficient solution techniques for high performance computing, which has begun to replace classical frequentist inference procedures developed years …

Hidden Markov Model

The hidden Markov model or HMM for short is a probabilistic sequence model that assigns a label to each unit in a sequence of observations. The model computes a probability distribution over possible sequences of labels and chooses the best label sequence that maximizes the probability of …

A statistical language modeling is a task of computing the probability of a sentence or sequence of words from a corpus. The standard is a trigram language model where the probability of a next word depends only on the previous two words. But the state-of-the-art as of writing is achieved …

I know it's a little late to make explicit my goals for this year, but think of it as my goals for the remaining three quarters of 2017. 2016 was the year of transitions as I left my first job to pursue my second in Korea. Moving from Boston to …

The \(k\)-nearest neighbors algorithm is a simple, yet powerful machine learning technique used for classification and regression. The basic premise is to use closest known data points to make a prediction; for instance, if \(k = 3\), then we'd use 3 nearest neighbors of a point in the test set …