Linear algebra is important in machine learning. It is in the heart of many machine learning algorithms such as Singluar Value Decomposition (SVD) and Principal Component Analysis (PCA). These are just few examples where linear algebra shines. If you want to know how dimensionality reduction works you ought to understand what matrices and vectors are, what are eigenvalues and eigenvectors, what inverse of a matrix stands for, what is covariance matrix is, and many other important aspects. A very brief introduction (in just 4 pages) into linear algebra can be found here. There exists a rife number of literature on linear algebra.

Gradient descent is a powerful algorithm that is being used in machine learning extensively. In this article we are going to demonstrate the power of this algorithm using simple example - polynomial regression. Keep on reading here.

If you face a problem of large datasets and how to store those on disk and later on use them to train the machine learning algorithms, dimensionality reduction can help a lot. PCA allows one to reduce the amount of features in the dataset with minimal lose of information (well at the end it is the engineer who controls how much information he or she is eager to lose during the transformation). Read more to discover how PCA works in practice.

SVD can be viewed as generalization to PCA. Similarly to PCA, SVD allows compressing the matrices with minimal loss of information (well, one can control how much compression he or she wants). In this article we show how to decompose matrices using SVD.

Understanding the mathematics behind algorithms is important for several reasons. One such reason is that the reader will have more clear picture of what the algorithm is doing under the hood and how to fine tune it. In this article, we will discuss the mathematics behind Support Vector Machine (SVM).

Naive Bayes classifier is the simple approach to classify categorical and continuous data using conditional probabilities under the assumption that these probabilities are independent for different features in the dataset. In this article we discuss how to use this type of classifier.

Often times there is a need to interpolate values which are spatially scattered. In this cases we can use methods like inverse distance weighting. However, this method does not take into consideration the continuity of the data (direction), but only distance between the point that we need to estimate and the rest of the known points. To take into consideration both distance and direction, it is best to use Kriging model. In this article we are going to discuss the mathematics of simple Kriging model.