prepare Streaming Data Science lecture

[emanueledellavalle]

prepare Streaming Data Science lecture to use when opening this part of the course. The lecture should call for Time-Evolving Analytics.

It should cover:

• standard DS assumption:
  • independent and identically distributed random variables
  • many pass algorithm
  • clear distinction between training and test
  • any task
• Time-Series Analytics: forecasting, when the independent distributed assumption does not hold
• Streaming Machine Learning: classification, single pass and prequential evaluation when the identically distributed assumption does not hold
• Continual AI: any task, when the identically distributed assumption does not hold
• Time-Evolving Analytics: when the entire i.i.d. assumption does not hold