Open lakillo opened 1 year ago
Is it possible that any other algorithm may be more suitable?
are there better models for time series data? XGBoost? autocorrelation function? History of data?
Deep learning probably overkill!
1) Can the technical team speak to domain experts / visit / connect with them to help with overcoming knowledge gaps? (we can probably rely on the postdoc for domain knowledge) 2) What are the expectations from using deep learning specifically? (recommend using a simpler approach) a) can we use ML over DL b) will there be domain expertise to help us extract the features? c) what compute facilities are needed
Deep learning.