Deep learning hackathon
Deep learning is developing rapidly, and the list of new breakthroughs and areas of its application is steadily growing ( image processing , speech recognition , reinforced learning , neuromachine translation , computing pharmacy 1 and 2 onwards). As a result, the world's largest IT companies (Google, Facebook, Baidu and many others) continue to actively introduce deep learning technologies, creating new jobs.
Meanwhile, the illusion persistently supported by journalists arises that the technology of deep learning is about to solve the problem of creating artificial intelligence [ 1 , 2 , 3, 4 ]. But the reality is that the range of unsolved problems will suffice for many more dissertations (see the presentation by Y. Lekun (Yann LeCun) on CVPR15, note by J. Schmidthuber (Jürgen Schmidhuber), post by B. Goertzel, already mentioned in the article on Habre work of J. Hinton (Geoffrey Hinton)). Aware of this fact, machine learning specialists strive to improve their skills; as an indicator, there were more than 600 applications for 100 places at the Yoshua Bengio summer school for deep learning this year.
Probably, not many of Habr’s readers had the opportunity to attend Y. Benjio’s school, however, it will be possible to gain experience and knowledge in deep learning in the course of an intensive weekly competition (hackathon), which will be held in Moscow in July. Hackathon participants will have the opportunity to listen to lectures from leading world experts, put their skills into practice and win prizes.
The event will be held at the Moscow Institute of Physics and Technology from July 19 to 25 and with the support of many companies (see pictures on the website), and senior students, graduate students, young scientists and IT professionals from all over Russia are invited to attend. From 30 to 50 people will be selected for participation in the hackathon, who will gather in teams of 3-4 people.
The main goals of the hackathon are: to unite people interested in machine learning and the problem of AI, to intensively obtain modern knowledge and practically apply them in an attempt to break world records.
The teams will be tasked with developing an algorithm that can independently learn to play any of Atari's computer retro games. This statement is determined by the fact that one of the most promising areas in the field of AI is the construction of neural network algorithms that can not only “recognize”, but also “understand”. In January this year, the DeepMind team (Google) published an articlein the prestigious scientific journal Nature, in which the best world result in developing a universal reinforcement learning algorithm was obtained using deep learning. This is just one of the first steps towards neural networks that can manage the environment around them and learn at the same time.
The participants will be helped by leading world experts who will give a series of lectures on the latest achievements in the field of machine learning at the Moscow Institute of Physics and Technology (live and remotely). See the list of lecturers on the hackathon's website, and what they will talk about, as a rule, already lies on their personal pages.
Sunday: gathering participants, organizational meeting, team building, setting up the environment, introductory lecture and the first night run of the calculation.
Monday-Friday: discussion of intermediate results, programming of new solutions developed in a team, lectures, launch of calculations.
After a week of development, the teams whose algorithms show the best results will participate in the final tournament “Hack the Game! Day ”, which will be held July 25 on Saturday at the site of the Polytechnic Museum at VDNH. And, of course, prizes will be awarded to the winning teams.
“Hack the Game! Day ”is planned as an open holiday for all those interested in games and artificial intelligence. On it, the best teams will present their results to visitors and tell you with what algorithms these results were obtained, and visitors can compete with the team bots.
Applications for participation in the DeepHack.Game hackathon are accepted until July 1, 2015.
All participants will be provided with accommodation and computing resources . But you need to bring your laptop with you.
Meanwhile, the illusion persistently supported by journalists arises that the technology of deep learning is about to solve the problem of creating artificial intelligence [ 1 , 2 , 3, 4 ]. But the reality is that the range of unsolved problems will suffice for many more dissertations (see the presentation by Y. Lekun (Yann LeCun) on CVPR15, note by J. Schmidthuber (Jürgen Schmidhuber), post by B. Goertzel, already mentioned in the article on Habre work of J. Hinton (Geoffrey Hinton)). Aware of this fact, machine learning specialists strive to improve their skills; as an indicator, there were more than 600 applications for 100 places at the Yoshua Bengio summer school for deep learning this year.
Probably, not many of Habr’s readers had the opportunity to attend Y. Benjio’s school, however, it will be possible to gain experience and knowledge in deep learning in the course of an intensive weekly competition (hackathon), which will be held in Moscow in July. Hackathon participants will have the opportunity to listen to lectures from leading world experts, put their skills into practice and win prizes.
The event will be held at the Moscow Institute of Physics and Technology from July 19 to 25 and with the support of many companies (see pictures on the website), and senior students, graduate students, young scientists and IT professionals from all over Russia are invited to attend. From 30 to 50 people will be selected for participation in the hackathon, who will gather in teams of 3-4 people.
The main goals of the hackathon are: to unite people interested in machine learning and the problem of AI, to intensively obtain modern knowledge and practically apply them in an attempt to break world records.
The teams will be tasked with developing an algorithm that can independently learn to play any of Atari's computer retro games. This statement is determined by the fact that one of the most promising areas in the field of AI is the construction of neural network algorithms that can not only “recognize”, but also “understand”. In January this year, the DeepMind team (Google) published an articlein the prestigious scientific journal Nature, in which the best world result in developing a universal reinforcement learning algorithm was obtained using deep learning. This is just one of the first steps towards neural networks that can manage the environment around them and learn at the same time.
The participants will be helped by leading world experts who will give a series of lectures on the latest achievements in the field of machine learning at the Moscow Institute of Physics and Technology (live and remotely). See the list of lecturers on the hackathon's website, and what they will talk about, as a rule, already lies on their personal pages.
Event Short Schedule
Sunday: gathering participants, organizational meeting, team building, setting up the environment, introductory lecture and the first night run of the calculation.
Monday-Friday: discussion of intermediate results, programming of new solutions developed in a team, lectures, launch of calculations.
After a week of development, the teams whose algorithms show the best results will participate in the final tournament “Hack the Game! Day ”, which will be held July 25 on Saturday at the site of the Polytechnic Museum at VDNH. And, of course, prizes will be awarded to the winning teams.
“Hack the Game! Day ”is planned as an open holiday for all those interested in games and artificial intelligence. On it, the best teams will present their results to visitors and tell you with what algorithms these results were obtained, and visitors can compete with the team bots.
Applications for participation in the DeepHack.Game hackathon are accepted until July 1, 2015.
All participants will be provided with accommodation and computing resources . But you need to bring your laptop with you.
Only registered users can participate in the survey. Please come in.
Are you interested in the topic of deep learning?
- 9% I work in the field of deep learning 10
- 70.2% I am going to start working in the field of deep education 78
- 20.7% I am not going to work in the field of deep education 23