Hi Naresh,

Truth is almost every model in AI or Machine Learning is a product of Linear Algebra, informed by Statistics and Probability . If you have a good understanding of these three subjects (Which frankly are relatively easy to grasp unlike Calculus), you'd be fine, no matter how complicated the algorithm may be. Note that deep learning is simply repeated, complex Linear Algebra, some times referred to as Matrix Algebra. In deep learning each node or neuron is simply divided into two parts, the first part computes the complex Linear Algebra, while the second part applies an activation function to the first part such as maybe Sigmoid, Tanh, Relu or Leaky-Relu activation functions.

Now about Calculus, if you already know high school calculus, just do a refresher. If you cannot remember or did not take calculus in school, you do not NEED to know calculus to do AI. If you know Linear Algebra, Statistics and Probability you can grasp any basics or Advanced knowledge of AI or Machine Learning without knowing beyond little about Calculus.

Where knowledge of Calculus is needed in AI is during Back-Propagation, which is when the model computes the Loss, then moves back from prediction to inputs one step at a time, calculating the derivatives or partial derivative outputs of each neuron in the deep neural network. This is the most complicated calculation in the entire field of deep learning or AI in general. So once you understand the general idea of back propagation and how derivatives simply mean slopes and how based on these derivatives the weights and biases of the entire neural networks are updated iteratively until the model accurately and confidently learns all the right parameters. Then you can simply gloss over or avoid going to learn Calculus, which in my opinion is more difficult to grasp than understanding the basics of Linear Algebra, Statistics and Probability.

I'm saying without knowing anything substantial about Calculus itself you can advance far into AI by simply understanding the entire forward propagation, loss computation and back-propagation steps to updating new weights and biases. Plus a good understanding of Linear Algebra, Statistics and Probability.

If you want to find out more take Andrew NG machine learning course or google ML crash course, then take the deep learning course from deeplearning.ai

Cheers.

------------------------------

Lawrence Krukrubo

------------------------------

Original Message:

Sent: Mon April 20, 2020 11:31 AM

From: Naresh Luthra

Subject: Level of Mathematical skill for AI

To what level does one needs to learn the mathematical skills like linear algebra / probability / statistics / calculus for AI/ML...Any resources which one can share on the above topics ...Thanks in Advance ...

------------------------------

Naresh Luthra

------------------------------