dtolpin/bocd

Bayesian Online Changepoint Detection

bocd's Issues

Interpretation of alpha, beta, kappa

Very nice implementation and very fast.

The setting these three parameters seems a bit mysterious to me. How would one
interpret them in order to properly instantiate a model?

Multi-dimensional input

Should this code work for series data that is multi-dimensional? I am testing it with x values that are 2D and getting the following error

File "/Users/carismoses/bocd/bocd/bocd.py", line 46, in update self.growth_probs[1:t + 2] = self.growth_probs[0:t + 1] * pred_probs * (1-H)
ValueError: could not broadcast input array from shape (2) into shape (1)

Just wondering if this is an issue on my end or if the bocd package doesn't support time series data of this shape.

Value of LAMBDA

The value of LAMBDA should be set to 100.0; otherwise an error may occur.