What to take as a token: a single word, an application, or the entire check?
Machine Learning / AI
Tell me about Mask R-CNN. What head is added on top of Faster R-CNN?
What is discretization (binning) of continuous features and why is it used?
How to solve the problem of missing necessary information in data?
Design a UGC content moderation system (text + images).
Is it possible to use additional features to improve the model?
What is meta-learning for AutoML?
What metrics are used to evaluate anomaly detection under imbalance?
What tasks do GNNs solve in molecular design (AlphaFold relatives)?
Is it possible to add input list sorting checks to protect against incorrect data?
What is selective search and why was it needed in R-CNN?
Is it necessary to add the remaining elements from a longer list after merging?
What types of augmentations are specific to detection (mosaic, mixup)?
What are node classification, edge classification, and graph classification?
How to merge two lists using a while loop with lists of different lengths?
Which tasks do GANs still solve better than diffusion models (real-time, low-latency)?
What is LambdaMART?
What are the manufacturing features when forecasting on thousands of time series?
How does a decision tree work?
In which tasks does CoT help, and in which does it break quality?