Distant-supervised Heterogeneous multitask learning for social - PowerPoint PPT Presentation

Distant-supervised Heterogeneous multitask learning for social event forecasting with multilingual indicators Liang Zhao George Mason University

What are Spatiotemporal Events? Week 47 Week 46 Week 45 Epidemics outbreak on Week 47 ending Nov 22, 2014 in southern region influenza Protests Civil unrest events on Mar 17, 2013 in Brazil

Open Source Indicators as the Social Sensor Protests on July 25, 2012, Mexico … … Tweet volume less than 10 Tweet volume larger than 10 A civil unrest event reported after July 25

Open Source Indicators as the Social Sensor 2013-14 Influenza Season Week 46 CDC flu activity map Flu tweets geographical distribution (reported on Week 47) (reported on Week 46) 1256 Flu tweets …

Challenge 1: Multilingual features 1. Must consider multilingual, because Moreover.. Countries with hundreds of languages • omit a group of people • Omit a language Cannot omit, even small ones • Social events can be triggered by any people 2. Too large dimension, too sparse feature vector Imagine a feature vector of a tweet: 10 nonzeros with 1M zeros.. 3. Few data for small language

Challenge 2: Cross-lingual semantic correlation 1. Features are highly semantically redundant One feature (https://www.profluentplus.com/blog/) 2. Features are correlated via multi-partite relationship (http://www.writeopinions.com/complete-multipartite-graph)

Challenge 3: Lack of language-wise supervision Zika outbreaks in Brazil No label on how much each group of language-speakers contribute (http://blogs.discovermagazine.com/science-sushi/2016/01/31/genetically- modified-mosquitoes-didnt-start-zika-ourbreak/) (http://www.foxnews.com/world/2017/12/18/mass-occupation- underscores-brazils-poverty-creates-angst.html)

Heterogeneous Multitask learning under distant supervision Task 1 (Language 1) Task 2 (Language 2) } Distant supervision Shared sparsity pattern Task 3 (Language 3) Word features Latent topics

Objective function Distant supervision: any language triggers, the whole triggers none language triggers, the whole not triggers Higher-level topic representation and transition matrix Shared sparsity patterns of Orthogonal constraint latent topics in different tasks Upper-bounded generalization error:

Optimization Equivalent problem: Alternating Direction Methods of Multipliers (ADMM) Solve Q: Dynamic programming Solve Θ and U : non-monotone spectral projected gradient descent Solve Z : second-order methods

Experiments

Distant-supervised Heterogeneous multitask learning for social - PowerPoint PPT Presentation

Distant-supervised Heterogeneous multitask learning for social event forecasting with multilingual indicators Liang Zhao George Mason University What are Spatiotemporal Events? Week 47 Week 46 Week 45 Epidemics outbreak on Week 47 ending

Multitask Learning Lei Tang Arizona State University Nov. 6th, 2006 Lei Tang Multitask

Not to Cry Wolf: Distantly Supervised Multitask Learning in Critical Care Patrick Schwab 1

Active Multitask Learning Using Both Supervised and Latent Shared Topics Ayan Acharya , Raymond J.

Extending Unsupervised Neural Image Compression With Supervised Multitask Learning David Tellez,

Consistent Multitask Learning with Nonlinear Output Constraints Carlo Ciliberto Department of

Learning frameworks Self-supervised learning: (Auto)encoder networks Supervised learning Network

Self-Supervised Feature Learning by Learning to Spot Artifacts Wonbin Kim Self-Supervised

Stacking for supervised learning Stacking for supervised learning Niall Rooney, NIKEL,

Machine Learning for NLP Supervised Learning Aurlie Herbelot 2019 Centre for Mind/Brain

PCA CS 446 Supervised learning So far, weve done supervised learning: Given (( x i , y i )) ,

Semi-Supervised Learning Maria-Florina Balcan 03/30/2015 Readings: Semi-Supervised Learning.

RECSM Summer School: Machine Learning for Social Sciences Session 1.3: Supervised Learning and

CSE 158 Lecture 1.5 Web Mining and Recommender Systems Supervised learning Regression

CSE 258 Lecture 1.5 Web Mining and Recommender Systems Supervised learning Regression

CSE 158 Lecture 1.5 Web Mining and Recommender Systems Supervised learning Regression

Machine Learning Lecture 6 Unsupervised Learning with a bit of Supervised Learning Clustering

Life in the Fast Lane: the confluence lens George Varghese, Microsoft Research I drive fast

Binomial Distribution Binomial Experiment 1 The same experiment is repeated a fixed number of

sizes of non-null hypotheses Jennifer Brennan, Ramya Korlakai Vinayak, Kevin Jamieson

CS70: Jean Walrand: Lecture 23. Conditional Probability: Review Conditional Probability: Pictures

Privacy Preserving Record Linkage Linkage Elizabeth Ashley Durham Health Information Privacy

Data Science in the Wild Lecture 6: Running Experiments Eran Toch Data Science in the Wild,

iec LING cole normale suprieure-Paris, IEC Ling February 28 2013 1 / 109 Foreword:

Supporting Drupal-as-a-Service Providing Tech Support to Drupal Devs Add speaker name here Kyle