Basic knowledge of simple learning in machine learning

Abstract : This article has a general knowledge of machine learning, a brief introduction to what machine learning can do, and how to do it. The following is the translation.

In the past few months, I have talked with many decision makers about artificial intelligence, especially machine learning. Several of these executives have been asked by investors about their strategies in Machine Learning and where they are using machine learning. So why is this technical topic suddenly becoming a topic of discussion on the board of directors of the company?

Computers should solve problems for humans. The traditional approach is to "write" the required programs, in other words, we teach algorithms that solve computer problems. The algorithm details the process of solving the problem, just like a recipe. Many tasks can be described by algorithms. For example, in elementary school, we learned about the digital addition algorithm. When it comes to running this algorithm fast and perfectly, computers are more capable than humans.

However, the process of solving this problem is limited. How do we identify a photo of a cat? This seemingly simple task is difficult to describe with an algorithm. Let us wait a moment and think about it. Even simple instructions (such as "four legs" or "with two eyes") have their drawbacks, as these features may be hidden, or the photos may only show a part of the cat. If we encounter a task of recognizing a leg or an eye, it is as difficult as recognizing a cat.

This is where machine learning shows its strength. The computer does not need to develop algorithms to solve the problem, but uses examples to learn the algorithm itself. We use samples to train computers. For the example of recognizing cats, we need to use a large number of photographs of cats to train the system (supervised learning). In this way, the algorithm evolves, matures, and eventually recognizes the cat on a strange picture.

In fact, in this case, the computer usually does not learn the classic program, and does not even learn the parameters in the model, such as the edge weights in the network. This principle can be compared to the learning process of our brain (which contains neurons). Like the brain, unlike traditional programs, this edge-weighted network is almost impossible to understand by humans.

In this case, the artificial neural network learning method called deep learning has been extremely successful. Deep learning is a special kind of machine learning, which in turn is a discipline of artificial intelligence, and is the main branch of computer science research. Back in 2012, the Google research team successfully trained a network of 16,000 computers and identified cats (and other object categories) from 10 million YouTube videos. They use deep learning techniques.

Many of the problems related to practice are more likely to belong to the category of "recognizing cats" rather than "digital additions", so it is difficult to solve these problems with human-written algorithms. These problems are often the identification of patterns in certain data, such as identifying objects in an image, text in a language, or fraudulent activity in transactional data.

Here is a simple example, let's take a look at predictive maintenance. Imagine that many sensors are sending data streams, and sometimes some machines fail. The challenge now is to learn the patterns of the data flow that caused the failure. Once you have learned this pattern, you can identify this pattern during normal operation to prevent potential failures.

Although the principle of machine learning is not new, it is currently highly sought after by the public. There are three main reasons for this: first, the availability of large amounts of data for application and training ("big data"); second, we now have huge computing power, especially in the cloud; third, a series of open source projects make each Individuals use more or less algorithms.

Machine learning is not a substitute for traditional programming, but a complement to traditional programming. It provides tools that enable us to solve many types of problems that have been difficult or impossible to solve so far. All in all, these have given us new opportunities, and existing systems are increasingly being transformed and added to machine learning capabilities.

A repeating operation following a certain pattern is a typical example. Suppose you have a computer program that you can use to access one hundred of its functions by clicking on a series of complex menus, but you only actively use some of these features every day. By observing the steps you usually click, the computer can learn to predict your next move, which will increase the efficiency of your operations. By "learning" these repetitive data and objects, computers can automate many of the steps, speeding up the process.

More examples can be found in each area: tailored learning materials for individual students (especially “large open online courses”, ie MOOC), early diagnosis of diseases, online marketing target groups, customer churn, automatic identification of data quality Problem, or match user's personal information through dating service.

Because Spark has many advanced tools, it (integrated in Hadoop) has become the leading big data framework in machine learning. Talend is also moving towards this goal, and through modeling can abstract a higher level. Modeling can reduce complexity while also weakening the dependencies between underlying technologies that are rapidly evolving and have high thresholds.

Only a few experts need to really understand the algorithmic details of machine learning. However, it is also beneficial if everyone can understand the concepts related to machine learning, such as knowing that machine learning is learning patterns from examples and can be applied to new data sets. In the end, machine learning expands the categories in which machines can solve problems, thereby automating: specifically by decision making. This is what computers have learned; it makes decisions about new data based on the knowledge accumulated from the training data. On the one hand, we can apply machine learning to our business or circle through automated decision making. On the other hand, we ourselves are a data source, and other machines can analyze us and optimize their own business.

All in all, I want to leave a picture of this: computers can now not only follow clear instructions (such as adding numbers), but also learn through examples (for example, by training examples to identify cat pictures). Depending on the problem to be solved, one method may be more appropriate than the other. However, when these two methods are combined in some way, they will eventually bring more opportunities for automation.

48V 200Ah Powerwall

Given that the cost, size and ease of installation of lithium ion battery are reducing, to rely on renewable solar energy and Solar Lifepo4 Battery can be an optimal option for any residential, commercial or industrial. The back-up battery should eliminate virtually all outages, and using Li-ion should last at least five years. UFO 48V 200Ah home Lithium Battery is designed with new lithium iron phosphate (LFP) battery technology. It is operating at 48 volt, which offers a power capacity of 9.6kWh.

Wall Mounted LiFePO4 battery Features

● Safe
UFO POWER has more than ten years experience in the lithium battery industry. We use reliable LiFePO4 batteries to ensure excellent product quality for you.

● Long Service Life
The Lithium battery has More than 6000 cycles, a longer life span of up to 15 years approximately. Deeper depth of discharge without decreasing in battery performance.

● Wide Compatibility

Can be equipped with UFO Self-developed Communication Protocols Conversion Module, which provides compatibility with 10 popular solar inverters on the market.

●Real-time Battery Monitoring
The UFO Powerwall is equipped with LCD Display Screen, making LiFePO4 battery status checking easily.

●Easy and Quick Installation
Each Powerwall battery is equipped with battery bracket and screws for easy installation.

48V 200Ah Powerwall,Powerwall Battery Pack,Fully Charged Battery,48V Powerwall Lithium Battery

ShenZhen UFO Power Technology Co., Ltd. , https://www.ufobattery.com