Thanks for watching my video! Want to learn more about Machine Learning? Check out my 30+ video playlist on Python & scikit-learn here: th-cam.com/play/PLTsu3dft3CWhSJh3x5T6jqPWTTg2i6jp1.html&si=UOuvLD9dFqWtiN1T
Thanks for watching! I hope you enjoyed the video. For more Data content you can subscribe to my channel, I share new videos every week. You can join to the Data Science discord server of the channel for asking questions, contributing to the discussions and get help from the text channels: discord.gg/BaVm6Rt4h8 For those looking to expand their skills, check out my Data Science courses on Udemy: Data Science Projects: www.udemy.com/course/data-science-projects-3/?referralCode=AEC736448BA104C3EC3F Data Analysis Interview: www.udemy.com/course/data-analysis-interview/?referralCode=3270B750A08BE82F7994 Python for Data Analytics: www.udemy.com/course/python-for-data-analytics/?referralCode=0782B89299FFF7561184 Machine Learning with Python: www.udemy.com/course/python-machine-learning-course/?referralCode=09455A4817E14D6B83D8 Python Programming: www.udemy.com/course/python-basics-i/?referralCode=3C8A77721CCEA802B372 Time Series Analysis: www.udemy.com/course/python-for-time-series/?referralCode=439816EE7E65C91F1B55 Natural Language Processing (NLP): www.udemy.com/course/python-natural-language-processing/?referralCode=AFFEAA30617CA01E6819 Happy learning!
Great demo and thanks for sharing. Trying to learn more about these hyperparameter tuning approaches. Do you have any suggestions for using optuna or other such approaches for optimizing sampling procedures in large imbalanced datasets with lots of diverse features/data types? I’m working with a multi-faceted LLM dataset and I’m trying to better understand the various distributions and it seems I need a more sound hierarchical stratification approach to sampling records, but I’m a little overwhelmed with fleshing that out. I’m basically trying to ensure that my train-val-test splits are balanced and reflect these nuances in ways a simple random sample will miss.
Thanks! Using Optuna to optimize a hierarchical stratification process is a great idea-it'll let you fine-tune your sampling strategy across different features. You might also look into more advanced stratification methods that go beyond random sampling to account for all those feature distributions. SMOTE, though usually for balancing classes, could potentially help augment underrepresented groups if that's part of the challenge too. Thanks for watching!
Thanks for watching my video! Want to learn more about Machine Learning? Check out my 30+ video playlist on Python & scikit-learn here: th-cam.com/play/PLTsu3dft3CWhSJh3x5T6jqPWTTg2i6jp1.html&si=UOuvLD9dFqWtiN1T
Goood.
Thanks. The best tutorial on optuna
Thanks for watching!
Thanks for watching! I hope you enjoyed the video. For more Data content you can subscribe to my channel, I share new videos every week.
You can join to the Data Science discord server of the channel for asking questions, contributing to the discussions and get help from the text channels: discord.gg/BaVm6Rt4h8
For those looking to expand their skills, check out my Data Science courses on Udemy:
Data Science Projects: www.udemy.com/course/data-science-projects-3/?referralCode=AEC736448BA104C3EC3F
Data Analysis Interview: www.udemy.com/course/data-analysis-interview/?referralCode=3270B750A08BE82F7994
Python for Data Analytics: www.udemy.com/course/python-for-data-analytics/?referralCode=0782B89299FFF7561184
Machine Learning with Python: www.udemy.com/course/python-machine-learning-course/?referralCode=09455A4817E14D6B83D8
Python Programming: www.udemy.com/course/python-basics-i/?referralCode=3C8A77721CCEA802B372
Time Series Analysis: www.udemy.com/course/python-for-time-series/?referralCode=439816EE7E65C91F1B55
Natural Language Processing (NLP): www.udemy.com/course/python-natural-language-processing/?referralCode=AFFEAA30617CA01E6819
Happy learning!
Thanks. I'll have to use in competition.
Thanks for watching!
Great demo and thanks for sharing. Trying to learn more about these hyperparameter tuning approaches.
Do you have any suggestions for using optuna or other such approaches for optimizing sampling procedures in large imbalanced datasets with lots of diverse features/data types? I’m working with a multi-faceted LLM dataset and I’m trying to better understand the various distributions and it seems I need a more sound hierarchical stratification approach to sampling records, but I’m a little overwhelmed with fleshing that out. I’m basically trying to ensure that my train-val-test splits are balanced and reflect these nuances in ways a simple random sample will miss.
Thanks! Using Optuna to optimize a hierarchical stratification process is a great idea-it'll let you fine-tune your sampling strategy across different features. You might also look into more advanced stratification methods that go beyond random sampling to account for all those feature distributions. SMOTE, though usually for balancing classes, could potentially help augment underrepresented groups if that's part of the challenge too. Thanks for watching!