Information Theory (IT) as a basis for Machine Learning (ML)

Information Theory (IT) as a basis for Machine Learning (ML)
.MP4, AVC, 1280x720, 30 fps | English, AAC, 2 Ch | 2h 8m | 1.15 GB

Instructor: Stephen Rayward
Solve the game Mastermind using ML mathematical techniques. These include: IT, Simulated Annealing and Markov chains.

What you'll learn

Automate Excel Solver for different trials – which in turn requires learning and using Named Ranges.
For Visual Basic for Applications (VBA) developers understand the VBA code required to automate Excel.
To understand how Metropolis-Hastings can be applied to identify probable solutions.
To understand Simulated Annealing.
To learn how to apply Simulated Annealing (together with Metropolis-Hastings to identify a probable feasible solution.
To ascertain how all these techniques are combined to create a Machine Learning strategy.

Requirements

Participants should have access to a computer with Microsoft Excel.
Strong Excel skills are not required but are encouraged.
Use of a Mac-based operating system is unsuitable.
The course uses two courses as prerequisites: 'Practical Introduction to Information Theory' and 'Geometric Introduction to Markov Chain Monte Carlo'.
The instructor aimed to make the course understandable without the prerequisites. However those who would like the most complete understanding of the course are encouraged to complete the prerequisites. However Practical introduction to Information Theory is mainly suitable to students with strong High school mathematics.

Description

Machine Learning is a modern branch of mathematics and data science. To some extent, Machine Learning is not well-defined. Here Machine Learning means an algorithm that can adapt to find better estimates of scenarios using available information.

Similarly Information Theory can be considered as a new branch of mathematics, and one approach of applying Machine Learning is to use Information Theory as the basis.

Hence the objective of the course is to explain the basis of a Machine Learning approach using Information Theory.

The game Mastermind is used as the focus. Hence a sub-objective is to demonstrate and explain how the game Mastermind can be solved using a Machine Learning (ML) approach.

The methodology links the following mathematical techniques:

Representation of a problem as a probabilistic system,
Identifying the solution to the probability system using Information Theory, and using as the software basis for this step Excel Solver.
Identify feasible solutions using Markov Chain Monte Carlo (MCMC) – particularly the Metropolis-Hastings algorithm (MH) together with Simulated Annealing.

Who this course is for:

This course will appeal to a wide variety of participants. The direct participants include the following:
Those learning the fundamentals of Machine Learning.
Those who are Machine Learning experts – who wish to broaden their knowledge of Machine Learning algorithms.
Those with a general interest in mathematics.
Those who are interested in the application of mathematics to games and puzzles.
Anyone completing a major in mathematics at University.
Mathematical researchers (particularly applied mathematicians and probabilists).
VBA developers and Excel users wanting to know possible diverse applications. Although the course does not go into depth in VBA.
High school mathematics teachers wanting to show diverse applications of mathematics.
Indirect participants will include those for which this course will later be used as a prerequisite. This includes engineers such as Mineral Processing Engineers (who have an interest in mathematical modelling).