Multivariate Gaussian Mixture Model | Intuition & Introduction | example in TensorFlow Probability

preview_player
Показать описание


In this video, we now want to extend the concept to Multidimensional Spaces. We therefore need to make use of the Multivariate Normal/Gaussian, which also requires some considerations with respect to the number of parameters.

In the last third of the video, we will then do a short implementation in TensorFlow Probability with scatter and contour plots.

-------

-------

Timestamps:
00:00 Introduction
01:04 Mixing Distributions
03:19 Weighting coefficients
04:00 Extension to more classes
04:50 Latent class assignments
06:32 Directed Graphical Model
10:34 Joint Distribution
12:22 Marginalizing the joint
13:04 Counting number of parameters
16:01 Three Different Forms
18:22 Shared Parameters
20:09 TFP: Intro
20:27 TFP: Categorical Distribution
20:54 TFP: Batched Multivariate Normal
23:38 TFP: Gaussian Mixture
24:16 TFP: Sampling
25:23 TFP: Contour Lines
28:25 Outro
Рекомендации по теме
Комментарии
Автор

Your video is gold. Appreciate it. Danke.

ythaaa
Автор

if the samples in my dataset are represented by real valued vectors. like {batch_size, vect_length] should i used mixture of multivariable or univariable?

bediosoro
visit shbcf.ru