Welcome to pay attention to us, AI tutorial, learning resources, fun popular science, paper interpretation, you want to see all here!

Recently, many students want to try to embrace AI, but they don’t know how to start. Among them, the core of artificial intelligence is Machine Learning, which is the fundamental way to make computers intelligent, and its application is all over the field of artificial intelligence.

EliteDataScience teaches you how to go from a beginner to a beginner in machine learning. Get on the bus, don’t change the coins, the bus is free!

Are you ready to teach yourself machine learning, but don’t know how to do it?

You don’t have to have a PhD or be a technical genius to get a world-class machine learning education for free. Whether you want to become a data scientist or use machine learning algorithms in your development, you can actually learn and apply machine learning faster than you think.

This article shows you a few steps on the road to machine learning so you don’t get lost.

Step one: What is machine learning

It is a good idea to understand what machine learning is and the basic concepts of machine learning before you start to study it.

Simply put, machine learning is teaching computers how to learn from data and then make decisions or predictions. For true machine learning, computers must be able to learn to recognize models without being explicitly programmed to do so.

Machine learning is an interdisciplinary discipline of computer science and statistics. It comes in many different names, such as data science, big data, artificial intelligence, predictive analytics, computer statistics, data mining

While there is a lot of overlap between machine learning and these fields, they should not be confused. For example, machine learning is a tool in data science that can also be used to process big data.

Machine learning itself is divided into several types, such as supervised learning, unsupervised learning, reinforcement learning and so on. Such as:

Mail operators are using supervised learning in machine learning to sort spam messages into bins; E-commerce companies classify consumers by analyzing consumption data, using unsupervised learning in machine learning; Augmented learning is used in driverless cars, where computers and cameras interact with the road and other vehicles to learn how to navigate.

For a primer on machine learning, check out some online courses. Ng’s introduction to Machine learning course is a must-see for anyone who wants to get a basic understanding of what the field is all about.

And Sebastian Thrun’s Introduction to Machine Learning, which provides a detailed introduction to machine learning, with lots of programming to help you consolidate what you’ve learned.

Of course, there is also a free column produced by Jizhijun, where you can not worry about the installation environment, directly into the simple machine learning training:

A concise tutorial on machine learning

These classes are free!

Presumably with machine learning in mind, we are in the knowledge preparation stage.

Step 2: Prepare knowledge

Without a basic knowledge base, machine learning can indeed seem daunting. To learn machine learning, you don’t have to be a math professional or a computer programmer, but you do need to have core skills in these areas.

The good news is, once you’ve done the basics, the rest is pretty easy. In fact, machine learning is basically the application of concepts from statistics and computer science to data.

The basic task of this step is to make sure that you don’t fall behind in programming and statistics.

2-1: Python programming for data science

You can’t use machine learning without knowing how to program it. Fortunately, here’s a free tutorial on how to learn Python for data science: Poke here.

Note: Jingllue jizhi adds three more resources:

01 Basic Rules

A complete guide to learning Python in data science from zero

And a summary of more than 40 Python learning resources

2-2: Knowledge of statistics for data science

A knowledge of statistics, especially Bayesian probability, is a basic requirement for many machine learning algorithms. Here’s a tutorial to learn about statistics in data learning: Poke here.

2-3: Math knowledge to learn

The research of machine learning algorithm needs some knowledge of linear algebra and multiple calculus as the foundation. Click here to get a free tutorial: Poke here.

Step 3: Go into sponge mode and learn as much as you can about the principles

The so-called “sponge mode” is to absorb as much as possible the principles and knowledge of machine learning as a sponge absorbs water. This step is somewhat similar to the first step, but the difference is that the first step is to have a preliminary understanding of machine learning, while this step is to master relevant principles.

Some of you might think: I don’t want to do basic research, so why do I need to master these principles, as long as I can use the machine learning toolkit?

It’s normal to wonder, but learning the basics of machine learning is important for anyone who wants to apply it to work. For example, when you apply machine learning, you may encounter these problems:

  • Data collection is a very time consuming process. You need to consider: What type of data do I need to collect? How much data do I need? And so on.

  • Data hypothesis and preprocessing. Different algorithms require different assumptions about the input data. How can I preprocess my data? Is my model reliable against the missing data?

  • Interpret model results. The idea that machine learning is a “black box” is clearly wrong. Yes, not all model results can be interpreted directly, but you need to be able to judge the state of the model and improve them. How do I determine if the model is over-fitting or under-fitting? How much room does the model have for improvement?

  • Optimize and debug the model. Very few people start out with an optimal model, and you need to understand the nuances and regularizations between different parameters. If my model overfits, how do I fix it? Should I combine several models together?

In order to solve these problems in machine learning research, it is essential to master the knowledge principle of machine learning. Here are two world-class machine learning courses that will definitely benefit you:

  1. This machine learning course at Harvard will help you understand the process from data collection to data analysis.

(Hint: this course works better with Professor Ng’s course.)

  1. Stanford’s machine learning course clearly explains the core concept of machine learning: Poke here.

There are also two reference books worth reading: Introduction to Statistical Learning and Fundamentals of Statistical Learning

Download the original English versions of the two books.

An Introduction to Statistical Learning

Elements of Statistical Learning

Note: Students who are not used to the original English can read the Chinese version of these two books.

Jizhi also suggested that you check out the machine learning forums on Reddit:

Machine Learning module 1

Machine Learning Module ii

Machine Learning section 3

Of course, the machine learning section on Quora is also interesting: Poke here.

Stroll forum not easy to see high order knowledge, you cannot stay in rookie stage all the time is not, want to upgrade it is necessary to see professional paper. A good place to go is arXive, a site that collects preprints of papers in physics, mathematics, computer science and biology.

Artificial Intelligence

Machine Learning

If it’s too much work to search for papers yourself, you can start an account on arxiv-sanity.com that will give you the latest arXive papers under its v-v-sanity.com hashtag.

Step 4: Targeted practice

Once you’re in Sponge mode, you should have a basic idea of machine learning, and then it’s time to get down to business. Hands-on is about enhancing your skills through specific, thoughtful hands-on actions. There are three goals for this step:

  • Practice the whole process of machine learning: collecting data, preprocessing and cleaning data, building models, training and debugging models, evaluating models.

  • Practice on real data sets: Develop your own judgment about what kind of data fits into what type of model.

  • Explore in depth: For example, in the previous step, you learned a lot about machine learning algorithms. In this step, apply different types of algorithms to the data set to see which works best.

Once this is done, you can move on to larger projects.

4-1 Nine basic parts

Machine learning is a very broad and rich field, with applications in almost every industry. With so much to learn, beginners can easily panic and lose sight of the big picture when faced with so many models.

Therefore, we can roughly divide machine learning into nine parts:

ML Holistic learning:

Optimization:

Data preprocessing:

Sampling and splitting:

Supervised learning:

Unsupervised learning

Model to evaluate

Integrated learning

Commercial application

4-2 Practical tools

For starters, we recommend using off-the-shelf algorithms so that you can spend your time familiarizing yourself with machine learning processes rather than writing algorithms. Depending on the programming language you’re using, there are two great tools:

Python scikit-learn: A poke tutorial

Caret: Poke tutorial for R language

4-3 Practice operation with data set

This step involves the actual operation of building and debugging the model with data sets, translating the theory you learned in the “sponge mode” phase into code. We recommend you get started with the data sets at UCI Machine Learning Repo, Kaggle, and http://Data.gov:

UCI Machine Learning Repo

Kaggle data set

DataGov data set

Step 5: Machine learning projects

Finally, the final step, which is very interesting. So far, we have completed the stages of knowledge accumulation, mastering basic principles, targeted exercises and so on, and now we are ready to explore the larger project:

The goal of this step is to practice applying machine learning techniques to complete end-to-end analysis.

Tasks: Complete the following items, from easy to difficult.

5-1: Titanic survivors forecast

Titanic survivor prediction is a popular choice for machine learning, and there are plenty of tutorials available.

Python tutorial 1

Python tutorial 2

R Language Tutorial 1

R Language Tutorial 2

5-2 Write the algorithm from scratch

We recommend that you start with some simple aspects: logistic regression, decision tree, k-nearest neighbor algorithm, etc.

If you get stuck in the middle, here are some tips to follow:

  • Wikipedia is a good source of pseudocode for common algorithms.

  • Take a look at the source code for some off-the-shelf ML toolkits for inspiration.

  • Divide the algorithm into several parts. Write the separation functions for sampling, gradient descent, etc.

  • Before we start writing the whole algorithm, let’s write a simple decision tree.

5-3 Pick an interesting project or field of interest

Actually, this is probably the best part of machine learning, you can use machine learning to implement your own ideas.

If you really can’t think of a good idea, here are eight fun machine learning practices for beginners: Poke here.

conclusion

If you follow these steps step by step, I believe you will eventually have a little success in machine learning!

We have 10 more tips for beginners in machine learning:

  • Set yourself learning goals and deadlines and do your best to meet them.

  • Lay a good learning foundation and master the basic theory.

  • Combine practice with theory, don’t focus on one aspect only.

  • Try writing a few algorithms yourself from scratch.

  • Think from multiple angles and find practical projects that interest you.

  • Think about what value each algorithm can generate.

  • Don’t believe the hype about ML in science fiction movies.

  • Don’t pay too much attention to online debates about ML knowledge.

  • Think more about the “in/out” of data and ask more “why” questions.

  • On the set of wisdom, the first time to upgrade themselves →→ set of wisdom

Finally, I wish you all success!

Note: The original text is in English, so most of the learning resources listed are in English. If you are worried that your English is not good, it doesn’t matter, because the learning ideas are the same, it is ok to look for Chinese learning materials at the corresponding stage.

Pay attention to jinglue wisdom nuggets home page, wonderful content no longer miss!