directory

  • What exactly is machine learning?
  • What are its practical uses?
    • Don’t try to learn everything before you start
    • Don’t take deep learning as your first lesson
    • Don’t collect too much data – Be aware of the timeliness of the data
      • The hardware conditions
      • Software selection
  • Write in the last

Many people want to take the train of artificial intelligence in the 21st century, constantly comply with the changes of the Internet era, and strive to win their own foothold in the field of the era of continuous innovation.

Also in the era of rapid development, the term artificial intelligence seems to let us no longer strange, accompanied by the new wave of machine learning field.

Perhaps many people know little about the new term ‘machine learning’, so today the big bad Wolf will come to talk with you about what kind of mysterious color ‘machine learning’ actually exists.

What exactly is machine learning?

In layman’s terms,

Machine Learning (ML) is a multidisciplinary interdisciplinary discipline, which involves a wide range of fields, including probability theory, statistics, approximation theory, convex analysis, algorithm complexity theory and many other disciplines.

It is specialized in studying how the computer simulates or realizes the human learning behavior, in order to acquire new knowledge or skills, and reorganize the existing knowledge structure to continuously improve its own performance. Thus, through a series of behavioral evolution, make the machine more convenient and lifelike for human use.

Machine learning is the core of ARTIFICIAL intelligence. It is the fundamental way to make computers intelligent. Its application is all over the field of artificial intelligence. The goal is to design and analyze algorithms that allow computers to learn automatically.

What are its practical uses?

So far, machine learning has been widely used in the Internet, medicine and other fields, such as: Data mining, computer vision, natural language processing, biometric recognition, search engines, medical diagnosis, detection of credit card fraud, securities market analysis, DNA sequencing, voice and handwriting recognition, strategy game or a robot is applied, the need of machine learning and development, to achieve higher efficiency of the application requirements.

Next, the big bad Wolf will briefly introduce the basic content and simple learning methods of machine learning.

As for machine learning algorithm, it is a kind of algorithm that obtains rules automatically from data analysis, and uses the rules to predict the unknown data.

Because learning algorithms involve a large number of statistical theories, and machine learning is particularly closely related to statistical inference, it is also called statistical learning theory. In algorithm design, machine learning theory focuses on achievable and effective learning algorithms. Many inferential problems are procedurally difficult, so part of machine learning research is developing approachable algorithms.



For machine learning, the Big Bad Wolf also has the following suggestions to share with you.

Don’t try to learn everything before you start

In fact, learning in many fields is the same, and almost every field does not exist independently, it more or less has a certain connection and interoperability with other related fields.

Therefore, when we study the knowledge of related fields, we do not need to learn all the contents involved first, which will inevitably cause a lot of time waste, and the efficiency of learning will be relatively low. The same is true for machine learning. In general, machine learning courses and books are:

Linear algebra: matrix/tensor multiplication, inverse, singular value decomposition/eigenvalue decomposition, determinant, norm, etc

Statistics and probability: probability distribution, independence and Bayes, Maximum likelihood (MLE) and maximum posteriori estimation (MAP), etc

Optimization: linear optimization, nonlinear optimization (convex optimization/non-convex optimization) and its derived solution methods such as gradient descent, Newton method, genetic algorithm and simulated annealing, etc

Calculus: partial differentiation, chain rule, matrix derivatives, etc

Information theory, numerical theory, etc



Finish for the average person, learning these knowledge often take a long time and energy, easy to give up, and the related knowledge is a tool which is not an end, our goal or have a certain knowledge of machine learning and practice, so these courses can completely in machine learning at the same time, targeted supplement and understanding, that have a purpose and less time-consuming.

Don’t take deep learning as your first lesson

Although most people learn “machine learning” for deep learning, it is not a good move to take deep learning as the first lesson of machine learning for the following reasons:

The black-box nature of deep learning is more obvious, and it is easy to learn without understanding

The theory/model architecture/techniques of deep learning are still in flux

Deep learning experiments have high hardware requirements and are not suitable for self-learning or using personal computers

So Learning Machine Learning can begin with the most basic introduction, the Wolf here recommend Wu En teacher Cousera Machine Learning courses, (the Machine Learning | Coursera)

Don’t collect too much data – Be aware of the timeliness of the data

Although machine learning has not been around for a long time, there is a lot of material out there, with gigabytes of material to download or watch. And a lot of friends have “collecting addiction”, even at a stroke to buy more than a dozen books, but to really use and not much.

Machine learning is evolving and changing fast. In the introductory period, it is recommended to “small and precise” selection of materials, choose recently published and good word of mouth books.



Here’s the big bad Wolf to talk about the preparation for machine learning

The hardware conditions

Another common question is whether you can use your laptop for machine learning. The answer is yes, most commercially available datasets can run in your memory.

In the introductory phase, we rarely use very large data sets, usually the largest is MNIST, which can be run on a personal laptop. Deep learning can also be done with gpus on Windows laptops, please don’t repurchase machines in the name of learning…

Software selection

When it comes to deep learning, the big bad Wolf goes first to Linux because it supports many learning models better (mainly the Library for deep learning). But even if you’re running Windows, you can use Ubuntu on a VIRTUAL machine to learn. This is sufficient for small deep learning models, and large deep learning is rarely run on local/PERSONAL computers.

As for the programming language, Python is the first choice, because Python has good extension support, and the major toolkits have Python versions. However, in certain cases, it is ok to choose R as the programming language. Other possible languages include C++, Java, and Matlab, but I personally don’t recommend them. After all, according to the trend of machine learning, the application of Python in machine learning is also being promoted further.

Write in the last

Although many people have said that the 21st century is the century of biology, but now the era of the Internet has already arrived, this is the century of artificial intelligence. If you understand machine learning, data analysis has its own charms.

Let everyone try to apply machine learning knowledge to their own field, eliminating artificial knowledge barriers. Only in this way can machine learning technology be implemented in more and different fields, thus feeding machine learning research itself.

Technology is changing fast, and we do not despise those who chase hot topics. However, in this impetuous era, no matter what direction you choose, the most important thing is the ability to think independently and the courage to discard the false and retain the true.

Therefore, you may still know very little after reading this article, but each piece of knowledge is endless to learn, I hope that you are not in a hurry to accept everything, but also do not have the stomach to reject everything. Slowing down, thinking, and making a plan that works for you is not just the right attitude to do scientific work. It is the attitude of each of us towards life!



Refuse the noise of the outside world, whether it is encouragement or ridicule, holding the belief that ‘long wind and waves will sometimes, directly hang the cloud sail to the sea’, unswervingly work hard, one day you will find that the key to the door of success has already been brought into the bag by you!

Like it and follow it,

Windows notebook using GPU for deep learning related materials can follow my wechat public number “Gray Wolf hole owner” reply “machine learning” to obtain!

Big bad Wolf is looking forward to progress with you!