Editor's note: Infinite Red Chief Technical Architect Gant Laborde will introduce you to machine learning through funny videos and approachable text.
If your knowledge of artificial intelligence and machine learning is a big question mark, then this article is for you. I will gradually increase your Awesomenessicityâ„¢ through inspirational videos and approachable text.
Sit back and relax. These videos are 5-20 minutes long. If these videos do not give you enough inspiration, you can stop reading the following text. However, if you find that you have read to the end of the article, then you have enough knowledge and passion to enter a whole new world. Which step you take depends entirely on you.
Understand why machine learning is so popular now
AI has always been cool. In "Pong", the mobile table tennis racket is AI, and in "Street Fighter", it is also AI that uses combos to beat you to the sky.
Atari's table tennis arcade game launched in 1972
AI can always be achieved through programmers' guesses, and programmers guess how something should behave. Interestingly, programmers do not have the talent for programming AI as we often think. Googling "epic game fails" (epic game fails), you can see a lot of AI and physical system failures in the game (sometimes you can see the failure of experienced human players).
Anyway, AI now has a new talent. You can teach computers to play video games, understand language, and even how to recognize people or objects. The tip of the iceberg of these new technologies comes from an ancient concept that has only recently gained enough processing power to be able to exist in scenarios outside of theory.
I'm talking about machine learning (Machine Learning).
You no longer need to come up with an advanced algorithm. You only need to teach the computer to come up with an advanced algorithm.
How did this happen? Algorithms are more not written, but multiplied. I did not use reproduction as an analogy.
Wow! This is a crazy process!
After the algorithm is implemented, how can we not understand how the algorithm works? The video below shows an AI that clears Mario. As humans, we all know how to play side-scrolling games. But AI's forecasting strategy is crazy.
Impressive, right? This idea is surprising, isn't it? The only problem is that we don’t know machine learning, and we don’t know how to integrate machine learning into video games.
Fortunately, Elon Musk founded OpenAI, a non-profit company that allows you to connect any AI to countless games/tasks with just a few dozen lines of code.
Why you should use machine learning
I have two good answers to why you should care about machine learning. First, machine learning (ML) is letting computers handle tasks that we have never let computers handle before. If you want to do something new, not only new to you, but new to the world, you can use ML to do it.
Second, if you do not affect the world, the world will affect you.
Now well-known companies are investing in ML, ML will change the world, we will wait and see. Thought leaders warned that we must not let this new era algorithm out of the public eye. Imagine what would happen if certain companies controlled the Internet. If we do not participate, technology will not belong to us. I think Christian Heilmann said very well in his speech on ML:
We can hope that others will only use this power in good faith. I am not optimistic about this gambler. On the contrary, I would rather be part of this revolution. you can also.
Ok, now I am a little interested...
This concept is useful and cool. We have already understood this concept at a high level, but what happened in the process of machine learning? How does it work?
If you want to get started directly, I suggest you skip this section and read the "How to Get Started" section directly below. If your motivation is to apply machine learning techniques, you do not need to read this section.
If you want to try to understand the working mechanism of machine learning, it is recommended to watch the video below, which uses classic machine learning handwriting as an example to show you the logic of machine learning.
The network is like a function. As the number of layers deepens, small pieces of data are chewed and abstract concepts are finally obtained. Adam Harley made an interactive interface where you can hand-write the numbers yourself to see if the neural network model can correctly recognize and the activation of the corresponding layers.
The above interactive interface shows the process of data flowing through the trained model, while the JavaFXpert/visual-neural-net-server project (code published on GitHub) developed by JavaFXpert visualizes the network training process.
The data set used in the demonstration is the iris data set (from 1936), which visualizes the process of backpropagating weights on a neural network.
I discovered this visualization tool through JavaFXpert's report on machine learning for Java developers. Even if you are not a Java developer, I recommend you to watch this report by JavaFXpert because it introduces a lot of machine learning concepts.
You can watch this report on YouTube: (Duration: 1.5 hours)
https://?v=I7GMyP6jdU0
Translator's Note: If you have trouble accessing YouTube, there is a method to get the video download address at the end of the article.
These machine learning concepts are really exciting! There are breakthroughs in the field of machine learning every day, so please start now.
How to get started
There are many resources for machine learning. I will recommend two routes.
Fundamental
Through this route, you will understand machine learning at the algorithmic and mathematical level. I know it sounds difficult, but it is cool to understand the details and code from the beginning.
If you plan to join the main force of ML and participate in the discussion, then this is the route for you.
I suggest you try the application of Brilliant.org (for anyone who loves science, this is a great application), and then take the artificial neural network course above. There is no time limit for this course. You can learn ML when you need to use your mobile phone to pass the time.
Note that this course requires payment after level 1.
You can take Wu Enda's machine learning course (Stanford online course) at the same time. This course is recommended in the JavaFXpert report mentioned above. In addition, Jen Looper also recommended this course to me.
Many people warn that this course is difficult. For some people, this means retreating from difficulties, but for others, this is the reason for studying this course in depth and obtaining certification.
The courses are 100% free. However, if you plan to obtain certification, you need to pay a fee for it.
There are many hands-on practices in these two courses. If you successfully complete these two courses, everyone will be impressed because they are not simple.
However, more importantly, if you do complete these two courses, you will have a deep understanding of the realization of machine learning, which will help you apply machine learning to change the world in innovative ways.
Speed ​​Racer
If you are not interested in writing algorithms, but want to use machine learning to create the next breakthrough website/application, you should skip directly to TensorFlow and the corresponding crash course.
TensorFlow is the de facto standard for machine learning open source software libraries. It has countless applications, and you can even use TensorFlow in JavaScript.
If taking a course is not your style, you are lucky. Today, you don't need to learn the essence of ML to apply it. You can effectively use ML as a service, using models trained by technology giants.
But I still remind you that this will not guarantee the security of your data (even your own security), but the ML service is very attractive.
If you are convenient to upload data to Amazon/Microsoft/Google, maybe using ML service is the best solution for you. I like to think of these services as introducers to more advanced ML. In any case, it is a good idea to start now.
Let us be creators
I would like to take this opportunity to express my gratitude to all those mentioned in between and the video author. They gave me a lot of inspiration during my introductory process. Although I am still a newcomer in the ML world, I am very happy to illuminate the way for others to embrace this amazing new era.
When you start to learn new things, communicating with others is essential. Without friendly faces, answers, and lively discussion areas, everything will be difficult. Being able to ask questions and get answers is very different. You can follow me (@GantLaborde) on Twitter, as well as the others mentioned earlier. The MachineLearning section of Reddit is also a good place for communication.
The following is a real example. I discussed high variance and overfitting with my friends on Twitter.
I hope this article inspired you and the people around you to learn ML!
Acrylic Hang On Privacy Filter
Acrylic Hang On Privacy Filter,Privacy Filter Screen Protector,Acrylic Hanging Anti-Glare Privacy Filter,Monitor Acrylic Privacy Screen Filter
Guangdong magic Electronic Limited , https://www.magicmax.cc