Essential Mathematical Concepts for Starting Your AI/ML Journey

Artificial intelligence has been the buzz word since the beginning of this decades. We are seeing businesses redefine their approach to meet the customer expectations by incorporating AI in the different fields of work. This involves training the talent with best knowledge of concepts and application of AI. One question which is often asked by people is how much mathematical knowledge is required for understanding the concepts and calculations used in AI.

In this article, we will discuss about the minimum mathematical concepts one should know to start learning AI/ML. Here are a few important concepts one should familiarize themselves with.:

1. Matrices(Linear Algebra): In AI/ML, input data is generally represented in the form of matrices. Basic matrix concepts include the number of rows, the number of columns, and how a matrix can be represented using these dimensions. for example: if the number of rows in a matrix is 3 and number of columns is 2, then the matrix can be represented as shown:

After, knowing the basics of matrix representation , should learn about matrix transpose , multiplication of matrices, dot product between matrices etc. In the image below, input layer contains the input data for the algorithm in the form of matrix. So we can see that to have a basic understanding of the any AI or ML algorithm, knowledge of basic matrix operations is indispensable.

2. Calculus: Another mathematical concept which is required for deep understanding of the AI/ML is Calculus. Knowledge of calculus will be highly useful to understand how a AI/ML model fine tunes itself and improves its performances. A very important concept of calculus called minima finding is used here by the name of gradient descent. In gradient descent, the loss function (difference between the actual and predicted value) needs to be reduce to minimum to make our model predicts come as close to the actual output. This is done by making small changes in the model parameters and observing the loss value. In the below image, we can see that the weights used for prediction in the model is initialized at random value and then the weight is changed by small values to observe the change in loss value.

In the below representation , we can see that ass we approach the minimum of the loss function, our line gets fit properly among the data points.

There our other calculus concepts like chain rule, partial calculus which are very easy to understand and implement. Other than the above concepts , topics like probability, linear algebra topics like vectors are important.