LinkedIn Inc.

{blong,lizhang}@linkedin.com

The objective of this tutorial is to provide in-depth and systematic introduction of large scale machine learning challenges, algorithms, and architectures with focus on information retrieval applications. First, we

will introduce fundamental aspects of large scale machine learning and typical challenges for large scale learning in information retrieval systems. Second, we will present principal algorithm framework for large

scale learning, which covers traditional distributed learning frameworks, such as distributed gradient descent learning, as well as state-of-the-art works, such as alternating direction method of multipliers and

Bayesian distributed learning. Third, we will further introduce commonly-used large scale machine learning algorithms for information retrieval, which fall into two categories: large scale supervised learning

algorithms such as classification, ranking and regression and large scale unsupervised learning algorithms such as matrix factorization and clustering. In this part, we will discuss general aspects for each category

of algorithms as well as practical implementations of specific representative algorithms, such as large scale logistic regression, large scale gradient boosting tree, and large scale latent factor learning. Fourth, we

will discuss and compare different architectures for large scale machine learning, such as Hadoop and Spark. Throughout the tutorial, concrete examples of large scale machine learning as well as case studies

from real-world applications, such as ads recommendation and Web search, will be provided for illustrations and discussion.

This tutorial would be appropriate for everyone attending CIKM
2013. No prior knowledge of large scale machine learning is
required. We will only assume basic knowledge in machine
learning methods and

information retrieval systems.

The tutorial will cover the following topics within both
practical and theoretical scope of large scale machine learning for
information retrieval.

- Introduction to large scale machine learning
- Examples of large scale learning in the field of information retrieval: content optimization, computational advertising, and web search
- Metrics to measure large scale machine learning system performance
- Classical large scale learning frameworks --- ensemble learning based approaches, distributed gradient descent learning
- Large scale response prediction for recommendation
- Alternating Direction Method of Multiplier for large scale logistic regression

- Parallel matrix factorization

- Large scale data analysis -- Bag of Little Bootsraps

- Comparison of different large scale learning architecture --- Hadoop and Spark
- Other open problems

Bo Long is a Staff applied researcher at LinkedIn Inc, and was formerly a senior research scientist at Yahoo! Labs. His research interests lie in data mining and machine learning with applications to web search, recommendation, and social network analysis. He holds eight innovations and has published peer-reviewed papers in top conferences and journals including ICML, KDD, ICDM, AAAI, SDM, CIKM, and KAIS. he has served as reviewers, workshops co-organizers, conference organizer committee members, and area chairs for multiple conferences, including KDD, NIPS, SIGIR, ICML, SDM, CIKM, JSM etc.

Liang Zhang is a Staff Applied Researcher at LinkedIn Inc. He obtained his Ph. D degree at Department of Statistical Science,Duke University in 2008. He worked at Yahoo! Inc. as a Scientist from 2008 to March 2012.

Liang has done many work and published several papers on applying statistical approaches to real world Internet applications where we usually find massive data. He also has years of experience of using Map-Reduce

and Hadoop system for his own Statistical research. Liang's research interests include recommender systems, computational advertising, statistical modeling and analysis for large scale data.