Is it possible to replace a 3x3 convolution with two consecutive 3x1 and 1x3 convolutions? What are the advantages and disadvantages?
Machine Learning / AI
How to decide on the loss function for a model from a business and data perspective?
What is the mathematical idea behind diffusion models? What are the forward and reverse processes?
What is fastembed and why is it convenient for prototyping?
How to perform cross-validation on time series?
Why not work directly with the source database?
Why choose the Kimball's star schema over the Operational Business Table (OBT)?
Tell me more about the Root Cause Analysis (RCA) project — what it is and how the system was implemented.
Is it clear what the company does (processing reviews on marketplaces and geoservices, analytics for partners)?
What risks does AutoML pose in terms of interpretability?
Tell us about yourself.
What types of sorting do you know and what are their time and memory complexities?
What is tree-of-thoughts?
What is SAC (Soft Actor-Critic) and why use maximum entropy RL?
What is a bottleneck layer and what is it used for?
Why is __len__ needed in Dataset? Why do we need to know the length?
Can linear regression be used in cases of sample bias?
What semantics should be reflected in batching for identification?
How to check for duplicates by attribute in an SQL table?