Time Series Modelling and State Space Models: Professor Chris Williams, University of Edinburgh

- AR, MA and ARMA models
- Parameter estimation for ARMA models
- Hidden Markov Models (definitions, inference, learning)
- Linear-Gaussian HMMs (Kalman filtering)
- More advanced topics (more elaborate state-space models, and recurrent neural networks)
#datascienceclasses

[@evocation12]

great explanations. like three touch downs in a single game

[@mannyk7634]

I agree.. it is a master piece

[@marxman1010]

Where can get the class presentation pdf file?

[@marxman1010]

What is the meaning of Xnj at 29:58? Xn at state j?

[@gopikrishnam8279]

I think it's an indicator variable which is 1 when X was j at nth step. To interpret this, compare this to standard markov chain. To get emission probability of j in state i, we'll count all the j's emitted when in state j divided by all observations in state i. Now when the states are hidden, we take all the observations j, and weight them by probability of being in state i at that time step, then normalize the sum by dividing by the sum of weights to get a probability. Hope that clarifies.

[@WahranRai]

too confused !

[@priapushk996]

Rule of lectures: try to impart one or two takeaway concepts. 90 minutes of sprawl ranging from first principles to current research breaks the rule.

[@mannyk7634]

he has many take away, it is very logical and gave great intuition. Your altitude is the problem

[@korrapatisrujan7943]

This lecture is highly dry as there is no effort put in to explain the intuition behind all the math

[@mannyk7634]

he gave many intuition especially in the beginning 10-15 mins