Tell us about a machine translation task. What is sequence-to-sequence?
Machine Learning / AI
How does MobileNet v2 differ from v1 (inverted residuals)?
What methods for combating overfitting exist and how do they work?
What is a skip connection and why is it needed (ResNet)?
What is the survivability function S(t)?
What methods are used for anomaly detection in images (PaDiM, PatchCore)?
What are A100, H100, B200, and their key architectural differences?
What is Triton (OpenAI) and where is it more convenient than CUDA C++?
What are IPO, KTO, ORPO as alternatives to DPO?
What is TGI (text-generation-inference) and its features?
What is gradient boosting?
What is stochastic depth and why is it used in EfficientNet?
What is neural-symbolic retrieval?
What are the features of online recommendation evaluation (CTR, conversion)?
What is HELM and what tasks does it cover?
What is fairness in ML and what definitions exist (demographic parity, equalized odds)?
List the metrics of binary classification model quality and their differences?
How does sparse MoE differ from dense MoE?
How to process data streams to calculate the median with limited memory?
What are Gaussian Processes and what kernel functions are used?