What is selection bias in click logs?
Machine Learning / AI
What is DistilBERT? What are the quality losses and speed gains?
What is RLHF? Describe the three stages of training.
How exactly was the embedder fine-tuned — LoRA or full fine-tuning?
What are the metrics for evaluating uplift models (Qini, uplift curve)?
How does TF-IDF differ from frequency representation and why is it needed?
What problems are there in detecting long-tail classes?
What is source separation (Demucs, Spleeter)?
What formula is used to minimize the distance between prediction and target in linear regression? How to take its derivative?
What is pose estimation (top-down vs bottom-up)? What is OpenPose?
What do you mean by document normalization?
What is test coverage and how to ensure it?
Tell me about Adam and RMSProp.
If we want to process multiple files in parallel, what is better to use?
What is spatial pyramid pooling?