Learn how to start deep learning.

In fact, for beginners, it is not recommended to learn the algorithm from the beginning, because it is meaningless to discuss the algorithm without business and data. First of all, we should lay a good foundation in programming and mathematics. Deep learning seems difficult. Learning according to the correct learning path can greatly lower the learning threshold and stimulate the fun of learning.

To put it simply, the learning route is as follows: first learn programming, mathematics and deep learning knowledge, then practice coding, take part in data science competitions if possible, and do more projects to practice practical ability.

As we all know, deep learning is a field in which theoretical algorithms and computer engineering technology are closely combined. What mathematical knowledge does zero-based Xiaobai need to master if he wants to develop into deep learning?

The first is linear algebra. In neural networks, a lot of calculations are matrix multiplication, which requires the knowledge of linear algebra. Inner product operation is also used to calculate cosine similarity of vectors, and various decomposition methods of matrices also appear in principal component analysis and singular value decomposition.

Followed by probability theory and statistics. Broadly speaking, the core of machine learning is statistical inference. Many giants of machine learning are statistical masters, such as Michael Jordan, Yang Lekun and Hinton. In addition, Bayesian formula and hidden Markov model are also widely used in machine learning.

Once again, calculus. This is one of the core knowledge in machine learning. Calculus is needed to calculate the gradient in gradient descent method or to deduce the error transmission in back propagation.

We know that deep learning is a field in which theoretical algorithms and computer engineering technology are closely combined. It needs a solid theoretical foundation to help you analyze data, and it also needs engineering ability to develop models and deploy services. Therefore, only by developing programming skills, machine learning knowledge and mathematics together can we achieve better results.

According to our learning experience, starting from a data source-even the most traditional machine learning algorithm that has been used for many years, we must first complete the whole workflow of machine learning, constantly try various algorithms, dig deep into the value of these data, thoroughly understand the data, characteristics and algorithms in the application process, and truly accumulate project experience, so as to master deep learning technology faster and more reliably.