Science is interesting. Science Club by MIL Team - a new format for working on scientific problems

Earlier this year, the Machine Intelligence Team launched a program for students and young scientists in the field of artificial intelligence - Science Club. Today we will tell you more about the program, about the participants and tasks, in particular about the topics and mentors in the field of natural language processing from the laboratory of neural systems and deep learning at MIPT .







What is Science Club?



Science Club is a new format for working on scientific problems, the goal of which is to give talented and young researchers the opportunity to work together with professional researchers. It requires some experience, time and desire from potential participants to achieve high results.



We invite you to learn:



  • , : , ;
  • SotA : , , ;
  • : β€” . , ;
  • : β€” , β€” , ;
  • : β€” AI;
  • : β€” , , .


Science Club . .


Science Club?



, .-...



Head of DeepPavlov.ai and Head of the Laboratory of Neural Systems and Deep Learning at MIPT. In the period 2017-2019. led the NTI project "NeuroIntellect iPavlov". In the 2019-2020 academic year, he was the scientific leader of the DREAM team, which participated in the Alexa Prize Socialbot Grand Challenge 3 from Amazon. The main organizer of the International Competition in Conversational Systems at the NeurIPS (Conversational Intelligence Challenge 1 & 2) conference, a member of the Expert Council of the Robotic Center and a member of the Expert Council of the European Commission Grant Committee for Future and Emerging Technologies, a member of the Scientific Council of the Polytechnic Museum.



Interests: neural network learning models, neurocognitive and neurohybrid systems, evolution of adaptive systems and evolutionary algorithms



Yuri Kuratov



Researcher at the Laboratory of Neural Systems and Deep Learning, MIPT. Winner of the 2017 NIPS Conversational Intelligence Challenge 2017 as part of bot # 1337, captain of the DREAM team in Amazon's Alexa Prize Socialbot Grand Challenge 3 competition in the 2019-2020 academic year.



Interests: language models, coreference resolution, BERT, question-answer models for SQuAD.



Vasily Konovalov



Researcher at the Laboratory of Neural Systems and Deep Learning, MIPT. Received M.Sc. in Computer Science. He is a lecturer at MIPT: Application of deep learning of neural networks for natural language processing tasks.



Interests: dialogue systems, question-answer systems, transfer learning



Alexey Goncharov



MIL Team Leader.



Interests: signal processing, semi-supervised, expert systems, time series clustering



One of these scientists can be your supervisor, a complete list can be found on the website . We are also working on attracting top leaders to the program, and if you have ideas for topics and published scientific papers, we are waiting for you in our ranks.


What tasks are set for scientists in Science Club?



All of our topics are in the field of AI. Some managers are involved in computer vision, others - natural language processing. Here, for example, we will share options for NLP topics from the laboratory of neural systems and deep learning at MIPT :



Transformer NLP ( DialoGLUE): intent recognition, slot filling, semantic parsing, dialog state tracking.



  • Intent recognition. : BANKING77 (single domain: banking), CLINC150 (10 domains), HWU64 (21 domains) .
  • Slot filling. ( ). : RESTAURANT8K, DSTC8.
  • Semantic parsing. TOR.
  • Dialogue state tracking. MULTIWOZ.


β€” https://arxiv.org/pdf/2009.13570.pdf, https://arxiv.org/pdf/2002.02450



(AutoML & NAS)



  • . , , β€” . , , , . , , .


(NLP & Conversational AI).



  • . , , , . .


β€” https://arxiv.org/pdf/1801.07243.pdf



  • . β€”- , .


β€” https://arxiv.org/pdf/2005.00613.pdf



  • , . .


Additional resources: - https://arxiv.org/abs/1503.02531 , https://arxiv.org/abs/1910.01108



A complete list of topics can be found on the website . Don't hesitate to pick a theme, as space is limited. If you have your own ideas, but need help implementing them and testing hypotheses, then you can find the right mentor in the community.


Ideas have already appeared. Where to go to submit documents?





To get started, you need to fill out the form on the site , indicating the desired start date and the topic you like. Next, we will contact you ourselves and give you a test task. After successfully completing it, we will be happy to talk with you, in person or in the zoom, to see if we are right for each other.



Don't be afraid to apply if you are not from MIPT. The program assumes the participation of any technical specialists. So wellcome)


Why is such a program needed? At conferences such as Sberbank Journey , the topic of leadership in AI and the role of young scientists from Russia in the development of this field is increasingly being raised. Therefore, the main goal of Science Club is not only productive scientific leadership, but the opportunity to create a world-class community, whose members write high-level articles (Q1) and speak at top conferences (NIPS). These are not just words, but a real achievable goal, on the way to which each participant can contribute to the development of AI on a global scale!



Do not put science on the back burner, become a part of it right now !



All Articles