Pavel Pleskov | Kaggle Grandmaster Interview | Kaggle Days

yeah so my name is five Swiss cough I'm a grandmaster currently and number three have a measure in mathematics and economics and I started Kegel like two years ago I quit my job as a quantitative researcher at high-precision term and I started to decide what I'm gonna do next salad do some Coursera courses including famous entry in G course and what they all lack was like practical skills some homeworks and I found those skills really easy to improve on Kegel so that's what I decided to do like full time and now I'm where I am actually I have a full-time job right now I work for a start-up called the point API it's an LP startup MIT based and I work in remotely so yeah it takes a lot of time and the rest I devote to Kegel and to traveling I participated to like 50 competitions and we really hard to pick one yeah there were a couple of first places but I think the most enjoyable was the one I got the third place it's about Russian language some Google IO P competition about Russian language and the reason is my favorite because I was travelling during this competition and stay in via couchsurfing and it was two days before the end I was sitting in front of my broken computer and there were kids running around me and screaming and I was like competing 12 hours per day in this setting so it was quite an experience be my heard libraries the one would be quite or CH and definitely it's like there is the interest in pytho she's really growing right now and I know people who even don't use numpy just pure fighters and this is in Cabo the second library would be like gvm I know of people back in the days like like eg boost but like GBM is better and like it was faster and and so on and the last one would be faster I definitely am really huge fan of Jeremy Javert and the work he did on building fast AI and promoting like the love of neural nets it's just incredible and I find it really useful even though some people complain about it but I liked it when I was beginning a library like a boosters library I think a lot of people were saying that by GBM is faster and you can actually get GPU functionality using just tip and not not building from the source and in this in this sense actually a cat boost is even better now because you can install like GPU version with it and it utilizes all GPUs out of the box and this is this is incredible so like a year ago CAD boost was really crappy but but now I think that's know is becoming number one yeah so there have been a lot of frustration when you complete a lot you definitely make some mistakes and one of the recent misfortunes what happened to me know we had a team of three and I've been a leader and I've chosen a submission which like I've so I've shown a submission like two hours before the end and I changed my mind to the other submission and we went from the tenth place and gold medal to the fiftieth place and silver medal and my team mates she was crying like the entire night and it was really frustrating for the whole team yeah that's really hard one I think like throughout the entire experience on Kegel and possibly six months it's youth don't give up I mean if you continue pushing you you will succeed I think that's the most most useful advice so yeah that's the top one so there are a lot of leakages in Kegel datasets I would say it's it's not it's not worse like comparing with people different platforms I think the most surprising for me is that for any kind of data you can you can augment your data so it's easy to imagine for images and even for NLP I know that Anthony was really surprised that you can submit your NLP dataset with translating that words like back and forth or finding synonyms and as recent Sandler competition showed you can even talk man table tabular data and this for me is incredible so I'm really impressed by rapid developments and NLP in let's say birth model and other other one was like that and recent developments in open AI models I think there was a lot of noise about that they didn't want to release the full train model because they thought could be damaging to like you know like in a sense of producing fake news I don't like totally agree with that but I really like those improvements in Adobe I think part is is the favorite part of Kegel days and like meeting people in person and in real life and hanging out with them that that's the most enjoyable part thank you you

Channel: Kaggle

Views: 9,492

Rating: 4.7064219 out of 5

Keywords: Kaggle, Kaggel, coffee chat, live-coding, live, learn, api, cli, python, data, data science, interview, questions, transfer learning, coding, networks, programming, technology, tech, machine learning, AI, artificial intelligence, coders, programmers, help, tutorial, projects, 101, rstats, stats, statistics, what is kaggle, how to, github, developer, kernels, datasets, data visualization, deep learning, sql, challenge, competition, whitehat, code, lesson, CS, Pavel Pleskov (ppleskov), kaggle days, grandmaster

Id: sF24xXGlOAI

Length: 6min 56sec (416 seconds)

Published: Thu Jun 13 2019