Front-end intelligence – the way of thinking transformation

What are the key issues?

The key question is: Do you do it or do the model do it? When you do it you have to think it through, when you model it you have to define it. Define the problem clearly and think clearly what the problem is? Many people get confused when they hear this… Put simply, a well-defined problem means: What problem in what area are you trying to solve? Problem thinking is: what problem in what area are you trying to solve? And how to solve it? Hearing this, you will have two questions: can you solve it? How to solve?

$ npm install -g @pipcook/pipcook-cli
Copy the code

Initialize a project:

$ mkdir pipcook-example && cd pipcook-example
$ pipcook init
Copy the code

Playground

If you are wondering what you can do in Pipcook and where you can check your training logs and models, you could start from Pipboard

$ pipcook board
Copy the code

You will see a web page prompt in your browser, and there is a MNIST showcase on the home page and play around there. If you want to train a model to recognize MNIST handwritten digits by yourself, you could try the examples below.

• pipeline-mnist-image-classification: pipeline for classific Mnist image classification problem.
• pipeline-databinding-image-classification: pipeline example to train the iamge classification task which is to classifify imgcook databinding pictures.
• pipeline-object-detection: pipeline example to train object detection task which is for component recognition used by imgcook.
• python-keras: example to use Python Keras library to train deep leraning network in js syntax and runtime.

See here for complete list, and it’s easy and quick to run these examples. For example, to do a MNIST

image classification, just run the following to start the pipeline:

$ node examples/pipeline/pipeline-mnist-image-classification.js
Copy the code

Documentation

Both Please refer to English | Chinese

Audience master: frighten? (/ yanke’) / ~ ╧╧

Ladies and gentlemen, Sir Mo Fang, I’ll yard a little bit more ha! In the way of “talking about the past and the present”, let’s take a look at the traditional r&d process and see why the problems that Da Vinci couldn’t solve with OpenCV can be solved with machine learning and intelligence. Back to the key question: What’s the difference between defining a problem and solving it? In the solution of OpenCV, the definition of the problem is how to use OpenCV algorithm to extract image, Text, Controller, we first need to define: what is the image, what is the Text, what is the Controller? However, to solve the problem with intelligent thinking, we just tell the model: this is image, this is Text, this is Controller, and this is annotated data. Like I said to the eldest son: you take blocks to throw brother is wrong, the results after a few days the eldest son took the ball to throw brother…… Even if I was breaking down inside, the problem was that I didn’t make it clear that it was wrong to throw things at other people, and that there was a lot more to explain when the older son was jealous of the younger one and that he was partially taking care of himself. The same is true for OpenCV to extract image, Text, and Controller. It is too difficult to figure out the situations that these elements appear in the design draft and the rules of their own feature description. The Da Vinci team has been thinking about it for more than two years. What can we do with machine learning and intelligence? Simply give the model samples of correct and incorrect answers and ask the model to use brute force algorithms — machine learning — to find ideas from the samples. So, here is the most essential difference is that we don’t want to know specific how to solve, as long as choosing the right model, each model has its own area, can check their framework, model related documents), tidy up the right answer, sample data, tell model, training model, the answer was completed! How to organize the correct answer? We created a SimpleCook platform for collecting, organizing, and annotating data sets. Collecting means finding raw data. In the imgCook project, we used front-end technology to generate controls for identification on the SimpleCook platform. Puppeteer is the key technology here. It uses Headless browser to render the page, and then labels the image areas corresponding to IMGCook, Text, and Controller using either program rule judgment or manual annotation. The data set — the production of the correct answer is over. How do you train the model? Remember the Pipcook front-end machine learning framework described above? One command to do it: See here for complete list, and it’s easy and quick to run these examples. For example, to do a MNIST image classification, just run the following to start the pipeline:

$ node examples/pipeline/pipeline-mnist-image-classification.js
Copy the code

Just follow the Pipcook tutorial and replace the sample data from the tutorial with your own data set. That’s all. To sum up, the key questions of intelligent thinking are: organize the right answer, select the model, train the model, and then what? That’s it! ~ ~

deterministic

First, the problem solving process of intelligent thinking is simple, and simple process represents certainty: there must be an answer. In fact, this is just like the confusion in life. Everyone has his own answer. Even if we tell the model the so-called correct answer, the model may not give us the expected answer to the unknown question after training, but the model will certainly give us an answer.

Secondly, from an experience point of view, as long as the problem is solved well in the industry, it should be solved well in your own field. For example, using an image classification model, it has been shown in rigorous experiments that the model can accurately identify a cat or a dog in a picture. If I tagged my cat and dog pictures and fed them to a model, the model would be trained to recognize other people’s pictures of cats and dogs, which is another level of certainty.

Finally, take a real example. For example, when you come to a new team, you can’t put all the faces and names together. After a period of practice, you communicate with each other every day, eat, drink and play together, and eventually you can remember everyone, and you can call any one of them by name. The same goes for the model, which at first couldn’t remember the faces of different people, but by labeling each person, each Angle, and each light condition with each person’s name, the model, with proper training, was able to call out each person’s name just as we do. That’s certainty.

evolutionary

When it comes to high availability on the server today, I talked to my colleagues at least 10 years ago about the “self-healing” of the system. Ten years later, few companies have come close to this goal. I’ll start by talking about the evolution of intelligent thinking, and I’ll end by sharing some of my explorations and thoughts about the “self-healing” of systems.

The evolution of intelligent thinking is reflected in the way we solve problems. In the past, we determined to understand the problem ideas before development, and realized the problem solving ideas in the development process, which is called Hardcode. When faced with problems that go beyond flexible design, we still have to write new code, constantly patching what we used to call “solutions.”

Intelligent thinking to solve the problem, we do not provide their thinking, model is extracted from the correct answer from their thinking, once encountered new situations we can put these things as a new feeding model, the model can self evolved, and similar problems to solve, this process is to realize the self evolution, The only thing we need to do is to form this evolutionary loop: the evaluation of model answers, the generation of positive and negative samples of new answers, and the construction of online training pathways.

The value of front-end intelligent thinking to solve problems lies in the certainty, robustness and evolution of problem solving. Ask, there is any programming language, technology, framework, scaffolding, tools… Can they give us that power?

Knowledge: Thought determines action

Having said all this, it’s no use if you don’t listen… What is the premise of listening? On reflection I think the first is to break the inertia. First of all, I admit that I am arrogant and opinionated as a programmer, because ever since I wrote my first function in Basic on Dos 6.22 in 1990, I printed out diamonds, circles, squares with “*” on the screen… I am sure I can write the whole world!!

Also that year, my father saved money and borrowed everything to buy me a 486 DX-100 computer, and I resolutely set foot on the road of “write the world”. Also in that year, I made every decision in my life: to go to technical secondary school, to drop out of technical secondary school after a semester to work, to be a waiter in a ballroom, to open a shop to sell videos and CDS, to close the shop for adult college entrance examination, to upgrade my law degree, to work as a salesperson in an ERP software company after graduation… All along the way, I made every decision on my own and no one would advise me, after all, I am the person who “writes the whole world”!!

Only by admitting my “headstrong” can I always remind myself to keep the “empty cup psychology”, which is an important reason why I learn things quickly, whether it is self-taught programming in the computer field to get to today, or self-taught electronic circuit design, embedded development, industrial design when starting a company…

Today, in the era of machine learning, new front-end intelligent thinking is needed to analyze, define and solve problems, and I happen to be able to clean myself up at any time and look at all new things and new technologies without inertia to find opportunities.

Join us

How to get internship Offer from Ali?

For details, please see: How to obtain The Internship Offer of Ali Front End? Ali P8 to teach you how to prepare for interview