Shandian Zhe: Probabilistic Machine Learning zhe@cs.utah.edu - - PowerPoint PPT Presentation

Shandian Zhe: Probabilistic Machine Learning

Research Topics:

• 1. Bayesian Nonparametrics
• 2. Bayesian Deep Learning
• 3. Probabilistic Graphical Models
• 4. Large-Scale Learning System
• 5. Tensor/Matrix Factorization
• 6. Embedding Learning

Assistant Professor, School of Computing, University of Utah

Applications:

• Collaborative Filtering
• Online Advertising
• Physical Simulation
• Brain Imaging Data Analysis

….

zhe@cs.utah.edu

Stochastic Nonparametric Event- Tensor Decomposition

Shandian Zhe, Yishuai Du

zhe@cs.utah.edu, yishuai.du@utah.edu

School of Computing, University of Utah

Tensor Data: Widely Used High-Order Data Structures to Represent Interactions of Multiple Objects/Entities

(user, item, online-store) (user, user, location, message-type) (subject, voxel, electrode) (user, advertisement, page-section) (patient, gene, condition) (user, movie, episode)

Tensor Decomposition

user item page purchase 100 25 35

• 23

21

• 100

25

• ..

.. .. 32 33

• user

item page

1/0

Traditional methods: Multilinear assumptions

Tensor Decomposition with Time

user item page purchase time-stamp 100 25 35

• 23

21

• 100

25

• ..

.. .. .. 32 33

• (i1, i2, i3)

=

f(

)

• …
• user

item page (user, item, page, week)

Stochastic Nonparametric Event-Tensor Decomposition

user item page time-stamp 100 25 35

• 23

21

• 100

25

• ..

.. .. 32 33

• index

time-stamps (100,25,35)

{s1, s2...sn}