I am currently studying the algorithm of machine learning, and I find that I use more mathematical knowledge as follows:

1, matrix correlation calculation, because machine learning processes multi-features and multi-samples, it inevitably involves matrices, and PCA and singular values are used in dimensionality reduction.

2, differential calculus, such as gradient direction, maximum and minimum.

3. Bayesian formula. Many models are based on Bayesian principles.

4. Statistical distribution, especially Gaussian distribution, is widely used.