Habr advise: how to win the NTI olympiad for schoolchildren in the Big Data profile

When I took my exams at the university, I always sat down to get ready for the school desk that was closest to the instructor and never went first, but missed 8-10 people ahead. Thus, I managed to “overhear” additional questions and, in part, answers to them.



And then I had an idea how to make tips on the Olympics for those high school students who read Habr (which means that they have really advanced brains). I found those guys who showed excellent results at previous NTI olympiads and asked them what additional materials could help to understand the topic better and would provide additional benefits at the final practical stage.



Accepting applications for the Olympiad is before October 22 , and there is still time to prepare.

Briefly about the NTI Olympiad

A group of students looked at what was going on with our education and decided to make an effort to fix it. What is happening? And the fact that the parents of many schoolchildren are not allowed to build a drone in the garage or take courses on building satellites, "because it is necessary to take the USE." So, our initiative comrades decided to create a movement, that talented schoolchildren will be able to get points for the Unified State Exam and enter higher education institutions not with the help of tests, but after working with their hands and head. And now the schoolchildren have an excuse for why he catched electrodes into the brain - “I am getting ready for the Olympiad”.

')

I think this is important and worthy. Yes, and I treat the competition with respect.

My reports on how the Olympiad was held in 2016

• All-Russian Engineering Olympiad for high school students: Space Systems
• All-Russian Engineering Olympiad for high school students: BigData and Intelligent Energy Systems

The organizers of the Olympiad themselves write about the requirements for the knowledge and skills of the participants:

• understanding basic concepts in machine learning and neural networks;
• programming in languages ​​that combine speed of development with the ability to use modern tools for working with data (for example, Python);
• the ability to quickly find solutions, to understand them and adapt to their needs on the Internet on specialized resources (stackexchange, etc.);
• the ability to work effectively with big data, write algorithms effectively; ability to test, debug;

And here are the recommended "official materials for preparation":

• Challenges for the 2015/16 academic year (PDF)
• Tasks at all stages of the Big Data and Machine Learning profile 2017 (PDF)
• Online course on Stepik: Preparing for the STI Olympiad (based on 2015/16 objectives)
• Remote Course on Stepik: Python Programming
• Online course on Stepik: Algorithms: theory and practice. Methods
• Online Course on Stepik: Discrete Structures
• Online Course on Stepik: Basics of Statistics
• MIPT Online Course on Coursera: Probability Theory for Beginners
• Yandex and MIPT Online Course on Coursera: “Learning from Marked Data”
• Online course of HSE and ShAD on Coursera: “Introduction to machine learning”
• Yandes and MIPT Online Course on Coursera: “Searching for a structure in the data”

How is the Olympiad for "Big Data":







During the final stage, the finalists are working to identify the specified information from the data arrays; for this, participants need to write their program using a set of texts for training. Workplace - a table, a laptop connected to the Internet.







The team works in “everyone at his own laptop” mode, from time to time discussing something with each other. To check the tasks, the moderator runs the program of the participating team with a set of random texts and evaluates the result of the data analysis.

Tips "experienced"

Anonymus 1111 advises.



To specifically drag the last stage in the “typewriter”, then something like this:







But in order to understand what a man says:

• You need to be able to jupyter * this is a general plus *
• It is very necessary to be able to Linux, the guys suffered directly. Here in 13 minutes you can at least understand what and how .
• And of course, so as not to sharpen for a specific task, because the next. if it is definitely not the same thing. Must be good at sklearn, numpy

Anonymus2222 advises.



General recommendations:

• start not with neural networks, but with simpler methods
• play with real data
• use Python in Jupyter Notebook

Specific resources:

• Introduction
• Standard course
• Standard library
• And a book on it
• Introductory task
• How does it work inside
• Collection of links
• Magazine

Anonymus3333 recommends.

• Hacker's guide to Neural Networks
• Understanding LSTM Networks
• Channel and specifically video:

About neurons







About tensorflow.







A simple math course on machine learning with vectors, terver and basic algorithms.







Advice from me personally.

We must try to be friends with the GoTo project with all our might, because there are also quality courses and teachers who are always ready to answer questions and hang out with students in Yandex on weekends. For the most cunning there are grants . In an extreme case, you can join the group in the VC and ask everyone for advice.



What would you advise high school students to improve their skills in the field of Big Data?