Lunedi 19

[filippozacchei]

Il nostro algoritmo e' "2-levele MCMC active learning"
Il loro algoritmo e' "MCMC guided active Learning"

Tasks:

• Implement MCMC guided active learning and assess performance on MEMS accelerometer tests case (M)
• introduce in the fodler the analytical banana and analytical funnel test case from UM-bridge (F)
• reason about online learning criteria in 2-level MCMC (F+M):

1. When to add the sample to the training dataset [ ]
2. When to stop the two level MCMC [ ]