How to become a Data Science Specialist: results of an open seminar at ITMO University
On May 16, a seminar on machine learning was held at ITMO University. The invited lecturer, head of the department of high-performance computer technologies at the Ural Federal University, Andrei Sozykin, spoke about the profession of a specialist in Data science and the directions of development of this field in the near future.
In today's material there are excerpts from interviews with the lecturer and a story about what the future data specialist needs to know and be able to do.
Flickr / Jer Thorp / CC
')
The Data scientist profession is relatively new, and not only for Russia, but for the whole world. Of course, not all tasks from the sphere of professional interests of a modern data expert appeared in recent years - some of them were previously solved by programmers, statisticians, and business analysts.
Moreover, the question of what exactly the Data scientist should know and be able to remain open: for example, on the website of the American Statistical Association, recently there was a debate about whether the emergence of a “data science” would bring death to statistics (and how closely related these disciplines are) and what is common to those who work in the positions of Business scientist, Data scientist, Data analyst and Statistician.
Of course, a large number of different terms and job titles cause some confusion. For example, Vincent Granville, an entrepreneur and researcher who developed an analytical direction in Visa, Microsoft, eBay and NBC, identifies as many as 16 different disciplines and professions that intersect Data Science in one way or another - from areas such as artificial intelligence and predictive modeling to professions like actuaries (in insurance) and quanta (in high-frequency trading). On the one hand, such a diversity can confuse a beginner, on the other - this is a clear sign that the future expert in Data science will definitely not remain without work.
Regardless of the name of this or that position, the data specialist is expected to have knowledge in several disciplines at once. Speaking at a lecture at ITMO University, Andrei Sozykin noted among the most important:
In order to "join" in this area, Andrei Sozykin recommends, in particular, the following courses:
According to Andrei Sozykin, it is possible to master the theory in about a year - especially if you are already studying for a specialty with a bias in statistics or IT. Medical or science background, work experience in the banking sector or insurance can also be most welcome.
Andrei emphasizes that it is important for the future specialist to have not only fundamental engineering knowledge, but also to understand the subject area in which the work will be carried out. In the end, one of the problems that large companies working with Big Data are facing now is the impossibility of effectively applying the results of research into practice.
Of course, a person with such a set of knowledge is a rarity. Therefore, Data science, as a rule, is not a single discipline, but a “command” discipline:
An unusual example of such a multidisciplinary approach is the work of Claudio Silva, Big Data and Data Science Specialist, a professor at the Polytechnic Institute and the Center for Urban Research and Progress at New York University. In 2015, he visited ITMO University for the first time and gave an interview about how Data science can be related to urban planning.
Claudio perceives the information that is generated in cities as “waste-free production”: Big data, created in the process of the work of numerous city services and enterprises, can serve the city as a blessing. For example, data specialists in New York have developed a product that allows urban road engineers to effectively use information about the movement of New York taxis.
According to Andrey Sozykin, the main directions for the development of Data science are hardware development to accelerate learning, creating more complex and accurate learning algorithms and building networks. An equally important task is to learn to better understand how the network “thinks” - it depends on how widely the development of Data science specialists will be applied in areas directly related to human life:
Note that at ITMO University work in the direction of Data science is done, in particular, by the Institute of High Technology Computer Technologies (NII NKT). About how the Institute's staff create models of events in places of mass gathering of people, analyze the mood of the crowd and assess public opinion according to social networks, we told in this material .
PS Already this Wednesday in the American Rapid City will take place the final of the World Championship on sports programming ACM ICPC 2017 (ITMO University is one of the leaders of the championship). Watch the live broadcast of the championship on May 24 and support our team!
In today's material there are excerpts from interviews with the lecturer and a story about what the future data specialist needs to know and be able to do.
Flickr / Jer Thorp / CC')
Data Scientist: analyst, mathematician, programmer
The Data scientist profession is relatively new, and not only for Russia, but for the whole world. Of course, not all tasks from the sphere of professional interests of a modern data expert appeared in recent years - some of them were previously solved by programmers, statisticians, and business analysts.
Moreover, the question of what exactly the Data scientist should know and be able to remain open: for example, on the website of the American Statistical Association, recently there was a debate about whether the emergence of a “data science” would bring death to statistics (and how closely related these disciplines are) and what is common to those who work in the positions of Business scientist, Data scientist, Data analyst and Statistician.
Of course, a large number of different terms and job titles cause some confusion. For example, Vincent Granville, an entrepreneur and researcher who developed an analytical direction in Visa, Microsoft, eBay and NBC, identifies as many as 16 different disciplines and professions that intersect Data Science in one way or another - from areas such as artificial intelligence and predictive modeling to professions like actuaries (in insurance) and quanta (in high-frequency trading). On the one hand, such a diversity can confuse a beginner, on the other - this is a clear sign that the future expert in Data science will definitely not remain without work.
Regardless of the name of this or that position, the data specialist is expected to have knowledge in several disciplines at once. Speaking at a lecture at ITMO University, Andrei Sozykin noted among the most important:
- Knowledge of mathematics, applied statistics and probability theory
- The ability to program (in particular in Python) and work with libraries for data analysis
- Understanding of actual methods of machine learning (which include not only deep learning)
In order to "join" in this area, Andrei Sozykin recommends, in particular, the following courses:
- Machine Learning (Coursera, course from Andrew Una)
- Neural Networks for Machine Learning (Coursera, course from Jeffrey Hinton)
- Data Analyst (Udacity, course prepared jointly with Facebook)
- Machine Learning Engineer (Udacity, course prepared jointly with Kaggle)
We also recommend our recent digest , which is entirely dedicated to the subject of Data science.
According to Andrei Sozykin, it is possible to master the theory in about a year - especially if you are already studying for a specialty with a bias in statistics or IT. Medical or science background, work experience in the banking sector or insurance can also be most welcome.
Andrei emphasizes that it is important for the future specialist to have not only fundamental engineering knowledge, but also to understand the subject area in which the work will be carried out. In the end, one of the problems that large companies working with Big Data are facing now is the impossibility of effectively applying the results of research into practice.
Of course, a person with such a set of knowledge is a rarity. Therefore, Data science, as a rule, is not a single discipline, but a “command” discipline:
This is a fundamentally multidisciplinary direction. [...] let's say someone programs well, someone at a very high level knows math, and someone understands the same banks, and together they give the result
- Andrey Sozykin
"Analytical Urbanism"
An unusual example of such a multidisciplinary approach is the work of Claudio Silva, Big Data and Data Science Specialist, a professor at the Polytechnic Institute and the Center for Urban Research and Progress at New York University. In 2015, he visited ITMO University for the first time and gave an interview about how Data science can be related to urban planning.
Claudio perceives the information that is generated in cities as “waste-free production”: Big data, created in the process of the work of numerous city services and enterprises, can serve the city as a blessing. For example, data specialists in New York have developed a product that allows urban road engineers to effectively use information about the movement of New York taxis.
It is important for us that all decisions made by city managers, engineers, architects follow the logic of the data so that they are not spontaneous or poorly weighed. We have the opportunity to more broadly look at how the city should develop, and we need to use it.
- Claudio Silva
According to Andrey Sozykin, the main directions for the development of Data science are hardware development to accelerate learning, creating more complex and accurate learning algorithms and building networks. An equally important task is to learn to better understand how the network “thinks” - it depends on how widely the development of Data science specialists will be applied in areas directly related to human life:
For us, it [the network] works in the so-called black box mode. We do not understand what is happening inside her and why she offers such options. In medicine, this is unacceptable, because in this area we must clearly explain and argue every action.
—Andrey Sozykin
Note that at ITMO University work in the direction of Data science is done, in particular, by the Institute of High Technology Computer Technologies (NII NKT). About how the Institute's staff create models of events in places of mass gathering of people, analyze the mood of the crowd and assess public opinion according to social networks, we told in this material .
PS Already this Wednesday in the American Rapid City will take place the final of the World Championship on sports programming ACM ICPC 2017 (ITMO University is one of the leaders of the championship). Watch the live broadcast of the championship on May 24 and support our team!
Source: https://habr.com/ru/post/329220/
All Articles