KDD 2012 was held in Beijing, China from 12th-16th August. Here is the main link to the conference: http://kdd2012.sigkdd.org/ . Complete list of accepted papers is available at: http://kdd2012.sigkdd.org/papers.shtml
Best Paper Awards
- Best Paper award : Searching and Mining Trillions of Time Series Subsequences under Dynamic Time Warping. Thanawin Rakthanmanon, Bilson Campana, Abdullah Mueen, Gustavo Batista,Brandon Westover, Qiang Zhu, Jesin Zakaria, Eamonn Keogh.
- Best Student Paper award #1 : Integrating Meta-Path Selection with User-Guided Object Clustering in Heterogeneous Information Networks. Yizhou Sun, Brandon Norick, Jiawei Han, Xifeng Yan, Philip S. Yu, Xiao Yu.
- Best Student Paper award #2 : Intrusion as (Anti)social Communication: Characterization and Detection. Qi Ding, Natallia Katenka, Paul Barford, Eric Kolaczyk, Mark Crovella.
- Best Paper in Industry Track : Bid Optimizing and Inventory Scoring in Targeted Online Advertising. Claudia Perlich, Brian Dalessandro, Ori Stitelman, Troy Raeder, Foster Provost.
Some of the papers covering topics of crowdsourcing, social networks, online learning, online auctions and advertisements :
- Adversarial Support Vector Machine Learning. Yan Zhou*, University of Texas at Dallas; Murat Kantarcioglu, University of Texas at Dallas; Bhavani Thuraisingham, University of Texas at Dallas; Bowei Xi, Purdue University
- Capacitated Team Formation Problem on Social Networks. A Majumder*, Bell Labs; Samik Datta, Bell Labs; Naidu KVM, Yahoo Labs, Bangalore
- Discovering Value from Community Activity on Focused Question-Answering Sites: A Case Study of Stack Overflow. Ashton Anderson*, Stanford; Dan Huttenlocher, Cornell University; Jon Kleinberg, Cornell University; Jure Leskovec, Stanford University
- Efficient and Domain-Invariant Competitor Mining. Theodoros Lappas*, Boston University; George Valkanas; Dimitrios Gunopulos, University of Athens
- eTrust: Understanding Trust Evolution in an Online World. Jiliang Tang*, ARIZONA STATE UNIVERISTY; Huiji Gao, ARIZONA STATE UNIVERISTY; Huan Liu, ARIZONA STATE UNIVERISTY; Atish Das Sarma, eBay Research Labs
- Factoring Past Exposure in Display Advertising Targeting. Abhimanyu Das*, Yahoo! Research; Sandeep Pandey, Twitter; Vijay Narayanan, Yahoo; Neha Gupta, University of Maryland, College Park
- Interacting Viruses in Networks: Can Both Survive?. Alex Beutel*, Carnegie Mellon University; B. Aditya Prakash, Carnegie Mellon University; Roni Rosenfeld, CMU; Christos Faloutsos,
- Joint Optimization of Bid and Budget Allocation in Sponsored Search. Weinan Zhang, Shanghai Jiao Tong University; Ying Zhang, Nankai University; Bin Gao*, Microsoft Research Asia; Yong Yu, Shanghai Jiao Tong University; Xiaojie Yuan, Nankai University; Tie-Yan Liu, Microsoft Research
- Learning from Crowds in the Presence of Schools of Thought. Yuandong Tian*, Carnegie Mellon University; Jun Zhu, Tsinghua University
- NASA: Achieving Lower Regrets and Faster Rates via Adaptive Stepsizes. Hua Ouyang*, Georgia Tech; Alexander Gray, Georgia Tech
- Online Allocation of Display Ads with Smooth Delivery. Anand Bhalgat*, University of Pennsylvania; Jon Feldman, Google Inc; Vahab Mirrokni, Google Research
- Online Learning to Diversify from Implicit Feedback. Karthik Raman*, Cornell University; Pannaga Shivaswamy, Cornell University; Thorsten Joachims, Cornell University
- SHALE: An Efficient Algorithm for Allocation of Guaranteed Display Advertising. Vijay Bharadwaj, NetFlix; Peiji Chen, Yahoo! Labs; Wenjing Ma, Yahoo! Labs; Chandrashekar Nagarajan, Yahoo! Labs; John Tomlin, opTomax Solutions; Sergei Vassilvitskii, Yahoo! Research; Erik Vee*, Yahoo! Research; Jian Yang, Yahoo! Labs
- Harnessing the Wisdom of the Crowds for Accurate Web Page Clipping. Lei Zhang ; Linpeng Tang ; Luo Ping ; Enhong Chen; Limei Jiao; Min Wang; Guiquan Liu
- Bid Optimizing and Inventory Scoring in Targeted Online Advertising. Claudia Perlich; Brian Dalessandro; Ori Stitelman; Troy Raeder; Foster Provost
WebSci 2012 was held in Chicago, USA from 22nd-24th June. Here is the main link to the conference: http://www.websci12.org. Complete list of accepted papers is available here.
Best Paper Awards
Papers from session of Social Computing and Collective Intelligence
AAAI 2012 was held in Toronto, Canada from 22nd-26th July. Here is the main link to the conference: http://www.aaai.org/Conferences/AAAI/aaai12.php/ . Complete list of accepted papers is available at: http://www.aaai.org/Conferences/AAAI/2012/aaai12accepts.pdf
Best Paper Award goes to: MOMDP: a Solution for modelling adaptive management problems. Iadine Chadès, Josie Carwardine, Tara G. Martin, Samuel Nicol, Régis Sabbadin, Olivier Buffet
Some of the papers covering topics of Machine Learning, Game Theory and Social Computing
- Automated Strategies for Determining Rewards for Human Work. Amos Azaria, Yonatan Aumann, Sarit Kraus
- Sembler: Ensembling Crowd Sequential Labeling for Improved Quality. Xian Wu, Wei Fan, Yong Yu
- Predicting Disease Transmission from Geo-Tagged Micro-Blog Data. Adam Sadilek, Henry Kautz, Vincent Silenzio
- Fairness and Welfare through Redistribution When Utility Is Transferable. Ruggiero Cavallo
- DUCT: An Upper Confidence Bound Approach to Distributed Constraint Optimization
Problems. Brammert Ottens, Christos Dimitrakakis, Boi Faltings
- Learning to Learn: Algorithmic Inspirations from Human Problem Solving. Ashish Kapoor, Bongshin Lee, Desney Tan, Eric Horvitz
- Knapsack Based Optimal Policies for Budget-Limited Multi-Armed Bandits. Long Tran-Thanh, Archie Chapman, Alex Rogers, Nicholas R. Jennings
- Learning Qualitative Models by Demonstration. Thomas R. Hinrichs, Kenneth D. Forbus
- A Convex Formulation for Learning from Crowds. Hiroshi Kajino, Yuta Tsuboi, Hisashi Kashima
- Computing Optimal Strategies to Commit to in Stochastic Games. Joshua Letchford, Liam MacDermed, Vincent Conitzer, Ronald Parr, Charles L. Isbell
- Optimizing Payments in Dominant-Strategy Mechanisms for Multi-Parameter Domains. Lachlan Dufton, Victor Naroditskiy, Maria Polukarov, Nicholas R. Jennings
- Strategic Advice Provision in Repeated Human-Agent Interactions. Amos Azaria, Zinovi Rabinovich, Sarit Kraus, Claudia V. Goldman, Ya’akov Gal
- Negotiation in Exploration-Based Environment. Israel Sofer, David Sarne, Avinatan Hassidim
- Approximately Revenue-Maximizing Auctions for Deliberative Agents. L. Elisa Celis, Anna R. Karlin, Kevin Leyton-Brown, C. Thach Nguyen, David R. M. Thompson
- Sequential Decision Making with Rank Dependent Utility: A Minimax Regret Approach. Gildas Jeantet, Patrice Perny, Olivier Spanjaard
- Discovering Spammers in Social Networks. Yin Zhu, Xiao Wang, Erheng Zhong, Nanthan N. Liu, He Li, Qiang Yang
- Optimal Auctions for Spiteful Bidders. Pingzhong Tang, Tuomas Sandholm
- Online Task Assignment in Crowdsourcing Markets. Chien-Ju Ho, Jennifer Wortman Vaughan
- Algorithmic and Human Teaching of Sequential Decision Tasks. Maya Cakmak, Manuel Lopes
- Time-Critical Influence Maximization in Social Networks with Time-Delayed Diffusion Process. Wei Chen, Wei Lu, Ning Zhang
- Quality Expectation-Variance Tradeoffs in Crowdsourcing Contests. Xi Alice Gao, Yoram Bachrach, Peter Key, Thore Graepel
- Dynamically Switching between Synergistic Workflows for Crowdsourcing. Christopher H. Lin, Mausam, Daniel S. Weld
AAMAS 2012 was held in Valencia, Spain from 4th-8th June. Here is the main link to the conference: http://aamas2012.webs.upv.es. Complete list of accepted papers is available here
Best Paper Awards
Some of the papers covering topics of Machine Learning, Game Theory and Social Computing
- Agents of Inﬂuence in Social Networks. Amer Ghanem, Srinivasa Vedanarayanan, Ali Minai
- DCOPs and Bandits: Exploration and Exploitation in Decentralised Coordination. Ruben Stranders, Long Tran-Thanh, Francesco Maria Delle Fave, Alex Rogers, Nick Jennings.
- Coordination Guided Reinforcement Learning. Qiangfeng Peter Lau, Mong Li Lee, Wynne Hsu
- Efficient Crowdsourcing Contests. Ruggiero Cavallo, Shaili Jain
- Optimal Incentive Timing Strategies for Product Marketing on Social Networks. Pankaj Dayama, Aditya Karnik, Yadati Narahari
- A Sequential Recommendation Approach for Interactive Personalized Story Generation. Hong Yu, Mark Riedl
- Learning in a Small World. Arun Tejasvi Chaganty, Prateek Gaur, Balaraman Ravindran
- Crowd IQ – Aggregating Opinions to Boost Performance. Yoram Bachrach, Thore Graepel, Gjergji Kasneci, Michal Kosinski, Jurgen Van-Gael
- Decision-Theoretic Approach to Maximizing Observation of Multiple Targets in Multi-Camera Surveillance. Prabhu Natarajan, Trong Nghia Hoang, Kian Hsiang Low, Mohan Kankanhalli
- Decentralized Bayesian Reinforcement Learning for Online Agent Collaboration. Luke Teacy, Georgios Chalkiadakis, Alessandro Farinelli, Alex Rogers, Nick Jennings, Sally McClean, Gerard Parr
- Repeated zero-sum games with budget. Troels Sørensen
- Combining Human and Machine Intelligence in Large-scale Crowdsourcing. Ece Kamar, Severin Hacker, Eric Horvitz
- A Truthful Learning Mechanism for Multi-Slot Sponsored Search Auctions with Externalities (Extended Abstract). Nicola Gatti, Alessandro Lazaric, Francesco Trovò
- Incentives for Truthful Reporting in Crowdsourcing(Extended Abstract). Ece Kamar, Eric Horvitz
Electronics Commerce(EC) 2012 conference was held in Valencia, Spain from 4th-8th June (EC 2012 link). It was co-located with AAMAS 2012 (AAMAS 2012 link). Here are some of the papers from EC12.
Best Papers Award
- Payment Rules through Discriminant-Based Classifiers. Paul Duetting, Felix Fischer, Pichayut Jirapinyo, John Lai, Benjamin Lubin and David Parkes.
Authors use SVM to come up with payment schemes which minimize ex-post regret and show them to be close to incentive compatible payments.
- Improving the Effectiveness of Time-Based Display Advertising. Daniel G. Goldstein, R. Preston MCafee and Siddharth Suri.
Authors shows that impact display advertisement has some diminishing returns. Hence displaying one ad for t time is less efficient compared to displaying two ads for t/2 time.
Some of the papers covering different flavors
Accepted papers in FOCS 2012 were out couple weeks ago and the list is available at: http://theory.stanford.edu/~tim/focs12/accepted.txt
Here are some of the papers related the areas of Machine Learning, Game Theory and Social Computing:
- A PTAS for Computing the Supremum of Gaussian Processes [ abstract ]
- Matching with our Eyes Closed
- A Tight Combinatorial Algorithm for Submodular Maximization Subject to a Matroid Constraint [abstract]
- A Tight Linear Time (1/2)-Approximation for Unconstrained Submodular Maximization [blog]
- Learning Topic Models — Going beyond SVD [abstract ]
- Active Property Testing [pdf]
- The Dynamics of Inuence Systems [abstract]
- Online Matching with Stochastic Rewards
- Finding Correlations in Subquadratic Time, with Applications to Learning Parities and Juntas[abstract]
- The Exponential Mechanism for Social Welfare: Private, Truthful, and Nearly Optimal [abstract]
- Optimal Multi-Dimensional Mechanism Design: Reducing Revenue to Welfare Maximizatio Equilibria
- Concave Generalized Flows with Applications to Market Equilibria [abstract]
I did some experiments on Mechanical Turk platform, to get an understanding of the platform and see the quality of data that flows in. Here are some details.
- You can register on this link: https://www.mturk.com/mturk/welcome
- You can get yourself registered as “Requester” to publish the HIT, and as “Worker” to complete the HIT. (HIT stands for Human Intelligence Tasks)
- It took me just 2 mins for registration, given I have Amazon account with US adress and US credit card. If you are out of US, it could be somewhat tricky on how to do this since currently this is only available for US residents.
- Money management of Requester is easy. I just brought 5 dollars credit online which is then used to complete the tasks I publish.
- The credit of Worker goes into account once the tasks have been approved by requester. Normally, the tasks get auto-approved in 1 week under default settings.
- HIT is designed in form of html page, which should take about 10 mins to finish. Then, you need to fill in some details like a) minimal qualification of work (at least X% approval rate in past), b) budget per task , c) total tasks in the HIT and so on. Overall, in 20 mins, you should be able to publish a HIT.
Recording the results
- Once the HIT is published, the results start flowing in quite instantaneously. I published my HIT of 100 tasks (details below) around 9pm Pacific Time and the HIT was completed in a timeframe of 5 hrs.
- The results can be viewed in almost real time and you do not need to wait for whole HIT to complete. The results can be exported to an Excel file.
Experimental HIT about Facebook
Answer a short survey related to your Facebook
1. Since how many years have you been using Facebook?
- Less than 1 year
- 1 – 2 years
- 2 – 3 years
- 3 or more years
2. Roughly, how many Facebook friends you have in following years? (enter 0 if you were not using Facebook at time of question)
- Number of friends in your Facebook, currently (June 2012)
- Number of friends in your Facebook, one year ago (June 2011)
- Number of friends in your Facebook, two years ago (June 2010)
3. Roughly, how many minutes in a day you spend on using Facebook
- Minutes spend on Facebook per day, currently (June 2012)
- Minutes spend on Facebook per day, one year ago (June 2011)
- Minutes spend on Facebook per day, two years ago (June 2010)
4. Roughly, from 1$ to 10000$, how much value would you assign to your Facebook account if someone asks you to sell your account? (enter 0 if you were not using Facebook at time of question)
- Value of your Facebook account, currently (June 2012)
- Value of your Facebook account one, year ago (June 2011)
- Value of your Facebook account two, years ago (June 2010)
5. (Optional) Please provide any comments you may have below, we appreciate your input
Experimental HIT results
- It is important to run your HIT for a small bacth so as to make sure that all the fields are correct. It is also important to analyze the workers response to ensure that they are not getting confused and frustated by your HIT.
- In total, I had batch of 100 tasks in the HIT. Here are some of the interesting responses.
- Feedback 1
- Years Usage = 3
- # Friends in 2012, 2011, 2010 = 368, 250, 150
- Time spent in 2012, 2011, 2010 (in minutes) = 30, 120, 30
- Value of Facebook account to sell in 2012, 2011, 2010 (in dollars) = 5000, 2500, 1250
- Optional Comments = I’m trying to quit
- Feedback 2
- Years Usage = 3
- # Friends in 2012, 2011, 2010 = 380, 300, 170
- Time spent in 2012, 2011, 2010 (in minutes) = 60, 180, 120
- Value of Facebook account to sell in 2012, 2011, 2010 (in dollars) = 8000, 10000, 8000
- Optional Comments = All of my friends including me, do not like timeline on facebook. It should be made optional
- Feedback 3
- Years Usage = 4
- # Friends in 2012, 2011, 2010 = 780, 500, 300
- Time spent in 2012, 2011, 2010 (in minutes) = 40, 30, 20
- Value of Facebook account to sell in 2012, 2011, 2010 (in dollars) = 10000, 8000, 2000
- Optional Comments = I enjoy facebook!
- Overall, it seemed that workers liked the HIT. Out of 100 workers, about 20 gave a positive feedback in “Optional Comments” section. This is useful to know since I can now deploy this kind of HIT on bigger scale to get more data.
- This was a variety HIT, I like it..
- It was interesting thinking back to how my facebook/my interaction with facebook has changed throughout the years.
- Very interesting and valuable study.
The purpose of this blog is to share ideas, discuss interesting papers and keep track of latest research in Crowdsourcing and Machine Learning.
To begin with, I will consolidate my blog posts from http://adishsingla.com/ into CrowdML blog.