How to Start Deep Learning and Artificial Intelligence

If you want to work in the field of artificial intelligence and deep learning, but do not know where and how to start, this article is prepared as a guide, based on the most common questions.


1.What is artificial intelligence?

Artificial intelligence, narrow intelligence and artificial general intelligence are divided into two.

Artificial general intelligence; is a set of software and hardware systems that are designed to be mathematically designed with the aim of visual perception, speech and voice recognition, movement, accounting and reasoning, which are designed mathematically by the human nervous system.

Artificial narrow intelligence is a narrow range of artificial intelligence systems developed to solve a specific problem and learn from data.


2.Is it difficult to learn artificial intelligence? Can I learn to myself?

No, it is not difficult. However, artificial intelligence is an area that requires serious time and work. In universities, scientific conferences, online education platforms, blogs and Youtube;

  • documents
  • videos
  • Open source applications
  • academic articles

You can also learn to yourself by following your level.


3.Where can I learn artificial intelligence algorithms and mathematics?

It is important to learn the basis of artificial intelligence. Otherwise, you are not a developer but an adaptor. Therefore, you should follow basic lessons such as artificial neural networks, machine learning, computational intelligence, deep learning.

Stanford Üniversitesi:

MIT (Massachusetts Teknoloji Enstitüsü):

Online Training Platforms:

Artificial intelligence, machine learning, deep learning etc. online training platforms where you can access the certified courses of online courses:


4.Which programming languages should I choose for artificial intelligence?

5 most common programming languages in data science:

  • Python (%57),
  • C/C++ (%44),
  • Java (%41),
  • R (%37), ve
  • JavaScript (%28)

Especially Python is the most preferred programming language because it is used in the background of many deep learning libraries, while the most preferred language for data visualization is R. Another advantage of the Python programming language is that it can be used for both academic and commercial applications.

The results of IEEE Spectrum’s research on the use of programming languages for 2018 are as follows:

5.What are the other alternatives for machine learning and optimization?

Of course, only programming languages in question 4 are not used. Other preferred environments after 5 most popular languages are:

  • Scala
  • Julia
  • Ruby
  • Octave
  • MatLab
  • SAS

6.What tools should I use to develop artificial intelligence?

On your own computer:

You should select the appropriate IDE (Integrated Development Environment) according to the programming language you are using. For example, if you are working in the most widely used Python language, you can choose Anaconda (Package Management Tool) and / or Visual Studio Code, Eclipse for Java.

Free cloud environment:

Microsoft Azure Notebook (CPU only) and Google Colab (it has GPU support) allow you to improve your application without requiring any installation.


7.They say the GPU is needed for deep learning, right?

True, but not necessarily. Because whether or not you need a GPU depends on your data set, the complexity of your model, and your time constraint.


8.For which problem should I choose / learn?

For this you need to examine how the methods have been applied to similar problems before. You have to do a literature review, but in an interview with Andrew Ng in Geoffrey Hinton’s “Heroes of Deep Learning” you should look at the paths in the literature and look at your paths. To master the theory will provide you with the necessary knowledge to design the appropriate model. Don’t forget to exchange ideas with your surroundings!

  • CNN (Convolutional Neural Networks): Object recognition and follow-up, style transfer, cancer detection, etc.
  • LSTM (Long Short Term Memory): Natural language processing, translation, chatbot, finance applications etc.
  • GAN (Generative Adversarial Networks): Synthetic data generation, fake face generation, style transfer, etc.
  • RL (Reinforcement Learning): Self-learning and low-intelligence artificial intelligence systems.

9.I don’t know Python, where should I learn?

You can learn from scratch or improve yourself by following related courses in online training platforms and examining the applications. It is an easy to learn programming language.

In order to develop deep learning applications with Python, you should look at the creator of the Keras library and the Google AI researcher François Chollet’s Deep Learning with Python.


10.What libraries (frameworks) should I use to develop machine learning, deep learning and optimization practices?

There are many libraries and APIs (Application Programming Interface) that have different features developed by various universities and companies for artificial intelligence, machine learning and deep learning.

According to the intensity of use, these libraries are as follows:

Alternatively, the following libraries may be used;

  • Tensorflow
  • Caffe
  • Caffe2
  • Torch/Pytorch
  • MxNet
  • CNTK
  • KNet
  • Theano
  • SciKit-Learn
  • Accord.NET
  • Spark MLIib
  • Azure ML Studio
  • Amazon Machine Learning

12. Where can I find the data sets?

If you can access it for free, you have a lot of data sets. The most well known are:

UCI Machine Learning Repository: Data Sets

Kaggle Datasets


13.Where should I start developing artificial intelligence for non-driver vehicles?

First of all, you should learn basic intelligence by following the answers of the questions about artificial intelligence and how to improve it. Following this, it will add value to following the courses “Deep Learning for Self-Driving Cars” opened in MIT and “Introduction to Self-Driving Cars“ in Udacity.


14.What are the applications of artificial intelligence in the field of finance?

One of the sectors in which artificial intelligence acts is finance. Especially in the field of finance, it is important to make predictions about the future and to keep the profit margin high. Market portfolio management, algorithmic purchase-sale transactions, fraud detection, credit / insurance commitments, user services, security 2.0, social media and press information from the press information based on the news and news analysis, sales related to produce recommendations such as making the machine can be done with the learning methods. and will continue to be popular in the future.


15.Can I use artificial intelligence to make a bitcoin prediction?

It is also possible to estimate cryptocurrency with machine learning or deep learning models. Although there are no promising studies yet, there are some studies about making bitcoin price estimation by machine learning.


16.What equipment are needed to design a deep learning model?

In fact, you only have one computer and internet connection is basically enough. So, everyone can design a deep learning model at home 🙂 But designing a model is possible by having scientific values. Therefore, you must have studied the design of a model. I would especially recommend that you use free cloud services in your studies, as there will be financial resources for students.


17.Will Artificial Intelligence take weapons and destroy humanity?

As Andrew Ng said, “Worrying about the fact that artificial intelligence is a bad super-intelligence is like worrying about the excessive population growth on Mars!” Therefore, studies are being carried out in order to regulate ethical values and legal rules on artificial intelligence in various countries. I would recommend that people understand the subject without any distraction in this area where people are foreign and carry out positive studies.


18.What is the gender of artificial intelligence?

Today, the subject can even question the gender of artificial intelligence. However, losing time with this type of speculation and manipulation produced for artificial intelligence, which is the field of today and the future, does not contribute to the development of the people of the country. Doesn’t it make you more excited to produce artificial intelligence instead of questioning the gender of artificial intelligence?


19.Why is Elon Musk trying to produce artificial intelligence if it is against artificial intelligence?

What Elon Musk opposes and perceives as a threat is artificial general intelligence.


20.What are the deficiencies and redundancies of artificial intelligence algorithms?

Especially in terms of deep learning; object recognition, face recognition, anomaly detection, natural language processing, translation etc. There is a lot of data in problem areas. While there is no need for data, there are no solutions for everyday problems other than computer games. In this process, there will be deficiencies in sub-branches of artificial intelligence, which will create new fields of work.

21.Can an artificial intelligence that acts like a human being be designed with the methods available today?

It is too early to tell a time to bring all the features, especially the reasoning of a person into a machine.

22.Are there no shortcomings for the neurons used in artificial intelligence to behave like humans?

Yes, there is certainly. The mathematical modeling of the human brain has not been fully understood as it is not yet fully understood. As improvements in the description of the human nervous system and brain functions occur, it will be possible to produce new solutions to artificial intelligence issues.


23.While neurons function in hundreds of structures and tasks in the human brain, human behavior cannot be created with only a few structures and tasks in the most advanced artificial intelligence. Is not it?

You’re right. Today, artificial intelligence systems are used to produce rational solutions to certain problems. To achieve integrity, learning needs to be self-evident and with little data.


24.Is data visualization necessary and how can I do it?

One of the important issues is data visualization because sometimes it can be one of the most decisive issues in understanding the problem. Understanding a problem shows how to get out of the way.


25.Is there a link between Artificial Intelligence and art?

For this, it is not possible to say yes or no, but I think it is an absolute connection between science and art, even if it is a relative subject. Therefore, products that may be perceived as art may arise. I can give a few examples for this: music production, visual production, data visualization with both visual and audio production, innovative product designs and so on.


Leave a Reply

Your email address will not be published. Required fields are marked *