Markov Models: Introduction to Markov Chains, Hidden Markov Models and Bayesian Networks

What is a MEMORYLESS predictive model? Markov models are a powerful predictive technique used to model stochastic systems using time-series data. They are centered around the fundamental property of “memorylessness”, stating that the outcome of a problem depends only on the current state of the system - historical data must be ignored. This model construction may sound overly simplistic. After all, if you have historical data why not use it to develop more complete and well-informed models? Surely, it would lead to more accurate predictions. However, when modelling time-series data where previous results are of limited relevance, a memoryless model delivers vast performance advantages. By considering only the present state, algorithms become highly scalable, stable, fast and, above-all-else, extremely versatile. Speech recognition is a perfect example - nearly all of today's speech recognition algorthms are built using Markov Models. In this book we will explore why a Memoryless predictive model can be so advantageous to the modern tech industry. We will take a look at fundamental mathematics and high-level concepts alike, extending our understanding of the subject beyond the simple Markov Model. You will learn... Foundations of Markov Models Markov Chains Case Study: Google PageRank Hidden Markov Models Bayesian Networks Inference Tasks