0:00:15 | without someone so the world that i'm going to present so you today is and |
---|
0:00:20 | extend personal for one of the sub subsystems so that the ice quality man |
---|
0:00:25 | submitted form needs the elderly and the false and fifteen challenge |
---|
0:00:30 | and the |
---|
0:00:32 | although it is focused on delivery a fifteen we believe that the key components or |
---|
0:00:38 | four |
---|
0:00:39 | in this of this paper they can be used in a much wider context |
---|
0:00:46 | so the first thing that we explore is that the how to more get the |
---|
0:00:52 | most of the key lda for |
---|
0:00:55 | and discriminant for discriminative training |
---|
0:00:58 | because nist ovaries that close the set identification task and it's going to train it |
---|
0:01:04 | should be the best |
---|
0:01:05 | so the most important thing is that we show that if we if we're |
---|
0:01:11 | and i u |
---|
0:01:12 | if we used plp parameters to projects i-vectors on the p lda latent subspace apply |
---|
0:01:17 | it discriminative methods in that subspace and then project them but |
---|
0:01:21 | then a week and improve the performance compared to just the baseline the one |
---|
0:01:26 | we use lda and then |
---|
0:01:28 | and maximum mutual information of their from on top of it |
---|
0:01:34 | and the second the |
---|
0:01:38 | important thing is that the we show how four |
---|
0:01:42 | take |
---|
0:01:44 | lre |
---|
0:01:45 | cost function |
---|
0:01:47 | approximated and use it as an objective function for a discriminative l the difference |
---|
0:01:52 | so that as a can see |
---|
0:01:55 | it is based on false acceptance rate and false rejection rate and all of them |
---|
0:01:59 | use indicator functions so we approximate those functions |
---|
0:02:04 | into continues functions and then |
---|
0:02:06 | where able to differentiate them |
---|
0:02:11 | and the of course of these method in general you can be used for any |
---|
0:02:16 | in a cost function in theory |
---|
0:02:21 | okay thank you for attention to the |
---|