How to perform Unsupervised Feature Selection using Supervised Algorithms
This article will provide you with a step-by-step process on how we came up with a new Machine Learning (ML) algorithm that performs unsupervised feature selection using any supervised algorithm of your choice such as XGBoost. Our goal is to provide you with insights into the process for developing a new ML algorithm.
How to correctly use TF-IDF with imbalanced data
If you have been employing TF-IDF with imbalanced data while setting some hyper-parameter that limits the number of features used, then you are probably hurting your Machine Learning (ML) model’s performance by supplying an incorrect feature-set. In this article, we will show you a nifty trick that can get you back on track!
An Unsupervised Neural Image Compression Algorithm
A lot of research has gone into developing deep learning based image compression algorithms that are mostly large neural networks trained in a supervised fashion. They require input in a particular size and work on only specific images. But what if you could create your neural compression algorithm that is unsupervised, can work with any image size and for arbitrary images? In this article, we develop just that!
How to invent your own new ML algorithm: A novel ensembling technique
Always wanted to create your own Machine Learning (ML) algorithm but didn’t know where to start? Well, you are in luck! This article not only helps you in understanding the thought process behind creating a novel ML algorithm but also develops one along the way. We create a completely new dynamic weighted averaging a.k.a. dynamic blending ensemble that outperforms its individual models as well as its static blending counterpart.
A simple technique to improve wind turbine misalignment detection
If a wind turbine is unable to align itself with the direction of wind flow it ends up generating sub-optimal amount of power leading to large financial losses. Traditionally, a lot of statistical analysis is required in order to detect turbine misalignment, potentially taking years for an exhaustive evaluation of multiple wind farms.
In this article, we provide a step-by-step intuition of developing a function that can accurately perform an exhaustive evaluation within seconds rather than years without the use of ML.