Скачать с ютуб Uber Eats Breakthrough in Recommendation Systems: Debiasing Position Bias with Deep Learning в хорошем качестве

Uber Eats Breakthrough in Recommendation Systems: Debiasing Position Bias with Deep Learning

#machinelearning #datascience #artificialgeneralintelligence #artificialintelligence #datatrek #recsys #recommendations #ubereats #uber #llm #medium #analytics #statistics UberEats Blog Link https://www.uber.com/en-IN/blog/impro... Overcoming Biases for a Better Recommender System    • Overcoming Biases for a Better Recomm...   A Guide to Model Calibration    • A Guide to Model Calibration | Calibr...   Other Recsys Videos on the Channel: Netflix's Calibrated Recommendations    • Tailored Streaming: Understanding Net...   Netflix's Unified Recommendation ML Model:    • Netflix's Unified ML Model: Deep Dive...   Evolution of Recommendation Systems | Instagram Pinterest Twitter    • Evolution of Recommendation Systems |...   Multi-Armed Bandit Startegies:    • Multi-Armed Bandit Strategies: Epsilo...   Building Scalable Query-Item Two-Tower Model based Retrieval System    • Building Scalable Retrieval System wi...   Chapters 0:00 - 1:26 Video Intro 1:27 - 6:23 Home Feed Ranking @ Uber & Different Statistical Biases 6:24 - 9:53 Position Bias in Uber Eats HomePage ~ Visualisation 9:54 - 15:16 Examination Model & Secondary Factors Affecting Position Bias 15:17 - 22:49 Handling Position Bias ~ Traditional Approaches & Novel Approach Proposed 22:50 - 24:22 Results & Conclusion The Uber Eats homepage plays a vital role in personalizing the user's food browsing experience. Recently, the research team at Uber published their work on mitigating position bias. Position bias is the phenomenon where users tend to order more from stores or items ranked higher, regardless of how relevant the store truly is to the user. Their work is novel, tweaking the model architecture to run on biased interaction data, effectively debiasing the conversion rate to extract the true conversion probability. The approach builds a deep learning CVR model with a position bias side tower, allowing simultaneous estimation of True CVR and Position Bias. Feature selection for each tower, along with regularization, is done carefully and cleverly to ensure each tower learns its own tasks without leakage from the other. This approach is an interesting way of tackling position bias compared to past techniques. It resulted in improved home feed recommendations for users and higher orders per user. Check out the detailed video I created around different biases that exist in recommender systems and the approach suggested by the Uber Eats team. Connect with Me Linkedin:   / mungoliabhishek81   Topmate: https://topmate.io/mungoli_abhishek Instagram:   / simplyspartanx   Twitter:   / mungoliabhishek   LinkedIn DataTrek:   / datatrek-channel   LinkTree: https://linktr.ee/abhishekmungoli #datascience #machinelearning #statistics #deeplearning #programming #python #datatrek #youtube #interview #interviewpreparation #interviewquestions #datascientist #dataanalytics #machinelearningengineer #datasciencejobs #datasciencetraining #datasciencecourse #datascienceenthusiast #career #careeropportunities #careergrowth #careerdevelopment #datascienceenthusiast #interviewing #ml #ai #datatrek #datascience #machinelearning #statistics #deeplearning #ai About DataTrek Series    • Introduction to DataTrek: Data Scienc...   Business Enquiries: [email protected] Find me on Instagram: www.instagram.com/simplyspartanx/ Music: www.bensound.com/royalty-free-music