James Koh

1 month ago | medium.com | James Koh

With diagrams and results obtained from actual numbersAnyone who is into Machine Learning would probably have seen an image similar to the one below when learning about regularization. Specifically, they would have been taught that by adding the L1-norm of the parameters to the cost function, and minimizing the combined loss, we would end up on a point along an axis, thus allowing us to effectively ‘drop a feature’ and simplify a model.

1 month ago | medium.com | James Koh

Maximum Likelihood Estimate (MLE) and Maximum A Posterior (MAP)You commonly see MLE and MAP in the first lesson of Machine Learning courses. How well will you be able to actually apply these if asked to formulate a problem? Let’s dive straight in. Consider the following scenarios. We can formulate these as a simple machine learning problem.

Feb 15, 2025 | medium.com | James Koh

Using args and kwargs to satisfy 7 different clientsJames Koh, PhD·FollowPublished inMITB For All·7 min read·--Imagine this. You are writing a Python programme which will be used by seven clients. Each client wants to use the function in a different manner, and all of them are important customers that your company wants to please. Suppose one of the functions is the computation of the Euclidean distance between two points. Let’s name this function dist.

Feb 9, 2025 | medium.com | James Koh

Maximum number of circles that can be packed into a rectangleJames Koh, PhD·FollowPublished inMITB For All·5 min read·--Consider a company that has to transport cylindrical materials in a lorry. What is the maximum number of pieces which can be stored? Assuming that the length of the cylinders are just slightly shorter than that of the lorry storage space, the problem can be simplified into a 2-dimensional problem like below. Cross-sectional view of stacked cylinders arranged by hexagonal packing.

Dec 9, 2024 | medium.com | James Koh

Entire training process with practical considerationJames Koh, PhD·FollowPublished inMITB For All·23 min read·--In this article, you would learn to set exploration and learning rates that would work in practically any state-action space, without worrying about any ‘wastage’ or ‘overshooting’ in training. This article assumes that you have basic knowledge of the RL concepts and terminologies, and basic Python programming skills such as using functions and classes.