Note: The following content is personal feelings and the study plan that I made for myself
I have been in contact with machine learning for 15 years. I spent a year literacy of commonly used machine learning algorithms. I remember that I used SAS and its cases and related books to implement the tools; I quickly forgot because I didn’t use it for work. At the end of the year and at the beginning of 2017, I "stir-fried" again, and I have almost forgotten it.
In the past few years, I have learned a lot of things, such as oracle, linux, and sas, but they are all "today's learning is for tomorrow's waste". The fundamental reason is that they have not applied what they have learned. To put it simply, they cannot be used in work; so learn well. The key to machine learning algorithms is that they can be used in work.
1. Belief
- Whether machine learning algorithms or great god algorithms are good, they are all "tools", and they are all created to solve practical problems.
- If you learn just to show off, then don’t learn it. It’s a waste of learning; if you learn that you can’t "landing" to solve practical problems, it's equivalent to not learning
2. Learning and memory
- Understand the principle and logic of each machine learning algorithm. You don’t need to study the principle deeply, and remember the logic after simplification
- Clarify what type of problem and what kind of scenario each machine learning algorithm is designed to solve
- Find out the characteristics of the sub-algorithms and extended algorithms of each machine learning algorithm, mainly to find out that it is to solve the problems of those special scenarios
- Clarify the overall classification structure of machine learning algorithms, which is helpful for overall understanding and memory
- Memory method: functional classification, what kind of problem to solve, typical case or work case
- In terms of python code implementation, it is only necessary to figure out the python library corresponding to the learning algorithm of each machine product, and there is no need to refine the specific code.
In a word, don’t go deep into the principles, remember the logic briefly, keep in mind what kind of problems to solve and typical cases
3. Application
- To clarify the overall structure of the machine learning algorithm is to "check the box" for practical problems, such as cooking pots and plowing the ground.
- Maybe you don’t necessarily need machine learning algorithms in your work, it may be simple OLAP, but when you have such problems, you should remember whether you can use them to solve the problem.
4. Easy to fall ill
I learnt machine learning algorithms easy to make mistakes: analysis for analysis, machine learning for pretense, and later when doing inventory forecasting, I found that the things I made with machine learning were not as good as operations managers can get by adding, subtracting, multiplying and dividing by pen. I think I’m ridiculous when I think about it.