For newbs like myself, the abbreviations used so far are: TBM: Transferable Belief Model; if my understanding is correct, in this video 2 TBM refers to the state transition model P(S_t|S_t-1) & the sensor model P(O_t|S_t). CPD: Conditional Probability Distribution
Correction for TBN it is 2TBN (Time Slice Bayesian Network). Where as Transferable Belief Model (TBM) is extension of Shafer work introduced by Smet's which oppose the current probabilistic approach in handling uncertainty.
@@goodn1051 That is actually a very bad approach, the sate of the art is either sentiment analysis as previously sugested or a hierarquical/mixture of experts that uses word2vec structures, HMM and NN., usually without sigmoid, using leaky/rely, since on large amounts of text it would suffer from exploding/vanishing gradients.
For newbs like myself, the abbreviations used so far are:
TBM: Transferable Belief Model; if my understanding is correct, in this video 2 TBM refers to the state transition model P(S_t|S_t-1) & the sensor model P(O_t|S_t).
CPD: Conditional Probability Distribution
God bless you
Correction for TBN it is 2TBN (Time Slice Bayesian Network). Where as Transferable Belief Model (TBM) is extension of Shafer work introduced by Smet's which oppose the current probabilistic approach in handling uncertainty.
stanford is always worth the time :) nice tutorial!
Best tut on HMM
Can someone say me her name?
Daphne Koller
Who else is watching this before big Buri exam?
What does the instructor say at 8:01, DDM?
I think it is DBN (i.e., Dynamic Bayesian Network)
Can this be used in detecting harassment or bully words in text?
It can be used, however, look and see how it has been done in the state of art!
look into sentiment analysis, it might work for your application
Better use a deep NN for that.....with a sigmoid function for output
@@goodn1051 That is actually a very bad approach, the sate of the art is either sentiment analysis as previously sugested or a hierarquical/mixture of experts that uses word2vec structures, HMM and NN., usually without sigmoid, using leaky/rely, since on large amounts of text it would suffer from exploding/vanishing gradients.
HMMMMMMMMMM