yt:channel:gGADKKGalfwSNbpSyM5ryg gGADKKGalfwSNbpSyM5ryg Yury Kashnitsky https://www.youtube.com/channel/UCgGADKKGalfwSNbpSyM5ryg 2011-08-17T20:14:08+00:00 yt:video:CdFv6NVYOO0 CdFv6NVYOO0 UCgGADKKGalfwSNbpSyM5ryg Yury Kashnitsky https://www.youtube.com/channel/UCgGADKKGalfwSNbpSyM5ryg 2022-11-21T10:42:26+00:00 2026-09-11T08:14:48+00:00 Blog "New Yorko Times" Welcome to my blog "New Yorko Times" https://yorko.github.io. This blog is about machine learning, mathematics, quantum computation, career development, programming, soft skills, popular science, and anything else I find exciting and (hopefully) can interestingly describe. My long-lasting nickname is Yorko (some African variation of Yury). And often times, something new is happening in my life. Hence, I’m titling the blog New Yorko Times. From time to time, I’d also write in my native language Russian. New posts are announced on LinkedIn https://www.linkedin.com/in/kashnitskiy and Twitter https://twitter.com/ykashnitsky. For Russian-speaking audiences, there's also a Telegram-channel "New Yorko Times" https://t.me/new_yorko_times yt:video:CPlYV_DryEo CPlYV_DryEo UCgGADKKGalfwSNbpSyM5ryg Yury Kashnitsky https://www.youtube.com/channel/UCgGADKKGalfwSNbpSyM5ryg 2022-01-31T14:57:44+00:00 2026-09-11T07:34:28+00:00 mlcourse.ai: free hands-on dive into practical Machine Learning In this video, we go through the https://mlcourse.ai self-paced roadmap and discuss how to approach the course in its self-paced mode: articles to read, lectures to watch, and assignments to crack. Course: https://mlcourse.ai GitHub repository: https://github.com/Yorko/mlcourse.ai Patreon to get bonus assignments: https://www.patreon.com/ods_mlcourse Russian version of the course: https://ods.ai/tracks/open-ml-course If you speak Russian, this is a preferred choice as there will be active sessions of the course (led by Peter Ermakov) – with assignments, Kaggle competitions, networking, etc. yt:video:A5brO93bRis A5brO93bRis UCgGADKKGalfwSNbpSyM5ryg Yury Kashnitsky https://www.youtube.com/channel/UCgGADKKGalfwSNbpSyM5ryg 2020-12-27T10:29:37+00:00 2026-09-11T08:42:38+00:00 BERT classifier fine-tuning with PyTorch, HuggingFace, and Catalyst. Part 4. Training with Catalyst The final part of the BERT fine-tuning tutorial https://github.com/Yorko/bert-finetuning-catalyst covers the actual training with Catalyst and GPUs. yt:video:StbFK_rp_rY StbFK_rp_rY UCgGADKKGalfwSNbpSyM5ryg Yury Kashnitsky https://www.youtube.com/channel/UCgGADKKGalfwSNbpSyM5ryg 2020-12-27T10:29:13+00:00 2026-09-10T07:02:48+00:00 BERT classifier fine-tuning with PyTorch, HuggingFace, and Catalyst. Part 3. HuggingFace BERT model In the 3rd part of the BERT fine-tuning tutorial https://github.com/Yorko/bert-finetuning-catalyst we try to understand the BERT classifier model by HuggingFace and dig into the code of the transformers library. yt:video:7zoNJV67dkA 7zoNJV67dkA UCgGADKKGalfwSNbpSyM5ryg Yury Kashnitsky https://www.youtube.com/channel/UCgGADKKGalfwSNbpSyM5ryg 2020-12-27T10:27:57+00:00 2026-09-10T11:16:40+00:00 BERT classifier fine-tuning with PyTorch, HuggingFace, and Catalyst. Part 2. Data preparation In the 2nd part of the BERT fine-tuning tutorial https://github.com/Yorko/bert-finetuning-catalyst we cover data preparation for training, from CSV files to PyTorch DataLoaders yt:video:fPDUcaLPu58 fPDUcaLPu58 UCgGADKKGalfwSNbpSyM5ryg Yury Kashnitsky https://www.youtube.com/channel/UCgGADKKGalfwSNbpSyM5ryg 2020-12-27T10:26:09+00:00 2026-09-10T20:05:52+00:00 BERT classifier fine-tuning with PyTorch, HuggingFace, and Catalyst. Part 1. Intro In the 1st part of the tutorial we describe the reusable BERT fine-tuning pipeline https://github.com/Yorko/bert-finetuning-catalyst I'm sharing the pipeline that I actually use at work (that's not a Kaggle Notebook anymore) which follows a modular approach, is easy to run and be reproduced. The follow-up parts of this tutorial cover: - data preparation for training, from CSV files to PyTorch DataLoaders - understanding the BERT classifier model by HuggingFace, digging into the code of the transformers library - running the pipeline with Catalyst and GPUs yt:video:JIU6WZuWl6k JIU6WZuWl6k UCgGADKKGalfwSNbpSyM5ryg Yury Kashnitsky https://www.youtube.com/channel/UCgGADKKGalfwSNbpSyM5ryg 2020-10-11T12:26:37+00:00 2026-09-12T14:54:50+00:00 Firing a cannon at sparrows: BERT vs. logreg There is a Golden Rule in NLP, at least when it comes to classification tasks: “Always start with a tfidf-logreg baseline”. Elaborating a bit, that’s building a logistic regression model on top of tf-idf (term frequency-inverse document frequency) text representation. This typically works fairly well, is simple to deploy as opposed to neural nets and that’s what's already deployed and working day and night while you are struggling with fancy transformers. In this presentation, we will go through a couple of real-world text classification problems and speculate on the reasons to resort to BERT as opposed to the good old tf-idf & logreg. In this talk, I share my experience finding a tradeoff between model performance and its ease of use and deployment in the context of multi-class text classification. Meanwhile, we will discuss a Catalyst text classification pipeline with HuggingFace. Slides: https://tinyurl.com/yytjbzz8 Code: https://www.kaggle.com/kashnitsky/distillbert-catalyst-amazon-product-reviews This talk part of the Catalyst track on DataFest 2020. Catalyst: https://github.com/catalyst-team/catalyst DataFest 2020: https://fest.ai/2020/ yt:video:9vvSp3YcE1k 9vvSp3YcE1k UCgGADKKGalfwSNbpSyM5ryg Yury Kashnitsky https://www.youtube.com/channel/UCgGADKKGalfwSNbpSyM5ryg 2020-09-07T06:26:04+00:00 2026-09-12T05:32:13+00:00 Прямая трансляция пользователя Yury Kashnitsky yt:video:FGuGg9F2VUs FGuGg9F2VUs UCgGADKKGalfwSNbpSyM5ryg Yury Kashnitsky https://www.youtube.com/channel/UCgGADKKGalfwSNbpSyM5ryg 2019-11-23T11:55:15+00:00 2026-04-29T00:51:15+00:00 How to jump into Data Science I'm Yury Kashnitsky leading mlcourse.ai https://mlcourse.ai – an open Machine Learning course by OpenDataScience (ods.ai). In this talk, I'll describe the learning path you need to step in to find your first DS job. Assuming that basic ML is covered (mlcourse.ai, Andrew Ng's course, or similar). I'll show you some typical questions that I like to ask at interviews myself. Slides https://www.slideshare.net/festline/how-to-jump-into-data-science yt:video:gyCjancgR9U gyCjancgR9U UCgGADKKGalfwSNbpSyM5ryg Yury Kashnitsky https://www.youtube.com/channel/UCgGADKKGalfwSNbpSyM5ryg 2019-11-07T12:39:49+00:00 2026-09-11T11:16:53+00:00 mlcourse.ai. Lecture 8. Part 2. Vowpal Wabbit Here we go on with Stochastic Gradient Descent and discuss Vowpal Wabbit library. During live coding, we train a linear model with several gigabytes of StackOverflow questions in just 30 seconds. Accompanying Jupyter notebook (messy, as is) - https://bit.ly/2Q1Ss1z Main site - https://mlcourse.ai Kaggle Dataset - https://www.kaggle.com/kashnitsky/mlcourse GitHub repo - https://github.com/Yorko/mlcourse.ai yt:video:DrohHdQa8u8 DrohHdQa8u8 UCgGADKKGalfwSNbpSyM5ryg Yury Kashnitsky https://www.youtube.com/channel/UCgGADKKGalfwSNbpSyM5ryg 2019-09-07T09:52:24+00:00 2026-09-10T19:51:19+00:00 mlcourse.ai Fall 2019 Live Session 0 Here we discuss the course roadmap, activities, what's new, some cool stories (maybe) etc. Slides https://www.slideshare.net/festline/mlcourseai-fall2019-live-session-0 yt:video:FrIW8ixKakw FrIW8ixKakw UCgGADKKGalfwSNbpSyM5ryg Yury Kashnitsky https://www.youtube.com/channel/UCgGADKKGalfwSNbpSyM5ryg 2018-12-23T11:04:54+00:00 2026-09-10T13:38:28+00:00 mlcourse.ai. Lecture -1. Outroduction Some directions to choose when you have covered basics of Machine Learning: - deep learning (cs231n) - theory of ML - more practice, Kaggle - first job in Data Science - pet projects Some words on the fall 2018 session of mlcourse.ai (thanks to Kirill Vlasoff and Andrew Lukyanenko for some insights/slides) Slides - https://bit.ly/2s0sjD7 Main site - https://mlcourse.ai Kaggle Dataset - https://www.kaggle.com/kashnitsky/mlcourse GitHub repo - https://github.com/Yorko/mlcourse.ai Patreon - https://www.patreon.com/ods_mlcourse (monthly support) Ko-Fi - https://ko-fi.com/mlcourse_ai (one-time support) yt:video:V5158Oug4W8 V5158Oug4W8 UCgGADKKGalfwSNbpSyM5ryg Yury Kashnitsky https://www.youtube.com/channel/UCgGADKKGalfwSNbpSyM5ryg 2018-12-09T20:40:36+00:00 2026-09-11T12:53:54+00:00 Topic 10. Part 2. Key ideas behind Xgboost, LightGBM, and CatBoost. Practice with LightGBM In this part, we discuss key difference between Xgboost, LightGBM, and CatBoost. Practice with logit, RF, and LightGBM - https://www.kaggle.com/kashnitsky/topic-10-practice-with-logit-rf-and-lightgbm Main site - https://mlcourse.ai Kaggle Dataset - https://www.kaggle.com/kashnitsky/mlcourse GitHub repo - https://github.com/Yorko/mlcourse.ai yt:video:g0ZOtzZqdqk g0ZOtzZqdqk UCgGADKKGalfwSNbpSyM5ryg Yury Kashnitsky https://www.youtube.com/channel/UCgGADKKGalfwSNbpSyM5ryg 2018-12-09T20:37:24+00:00 2026-09-11T02:06:11+00:00 Topic 10. Part 1. Gradient boosting basics In this video, we cover fundamental ideas behind gradient boosting, the versatile high-performing machine learning algorithm. Main site - https://mlcourse.ai Kaggle Dataset - https://www.kaggle.com/kashnitsky/mlcourse GitHub repo - https://github.com/Yorko/mlcourse.ai yt:video:_9lBwXnbOd8 _9lBwXnbOd8 UCgGADKKGalfwSNbpSyM5ryg Yury Kashnitsky https://www.youtube.com/channel/UCgGADKKGalfwSNbpSyM5ryg 2018-12-09T20:36:02+00:00 2026-09-11T12:02:07+00:00 Topic 9. Time series analysis Here we discuss foundations of the ARIMA forecasting model, which is accurate, useful for small time series prediction as well as it's important for understanding time series. We also briefly discuss the Facebook Prophet approach which is more suitable for scalable predictions. Slides - https://www.slideshare.net/festline/time-series-forecasting-with-arima-125447109 (by Evgeniy Riabenko) ARIMA example - https://www.kaggle.com/kashnitsky/topic-9-time-series-arima-example Main site - https://mlcourse.ai Kaggle Dataset - https://www.kaggle.com/kashnitsky/mlcourse GitHub repo - https://github.com/Yorko/mlcourse.ai