Data Science Lab

Introduction

A fundamental challenge in data science and big data analytics is to learn from data, behaviors, problems and systems that are not independent and identically distribution (non-IID or non-.i.i.d. for short). Non-IID learning uncovers both explicit and implicit (hidden, latent) coupling relationships and heterogeneities between attributes, attribute values, objects, views, sources, modalities, tasks, classes/labels, groups/clusters, methods/models, evaluation measures, and results, etc. This goes beyond classic theories, tools, and systems in statistics, information theories, artificial intelligence, machine learning, data analytics, pattern recognition, image processing, signal processing, data processing, and evaluation, etc., which typically take a IID (i.i.d.) assumption, and may result in partial, misleading or incorrect understanding of real-life challenges, complexities and intelligences.

This special session aims to demonstrate some state-of-the-art thinking, designs, applications, results, and discussions on non-IID learning and learning from non-IID data, behaviors, problems, and systems.

Aims and scope

Learning from big data is increasingly becoming a major challenge and opportunity for big business and innovative learning theories and tools. Some of the most critical challenges of learning from big data are the uncovering of the explicit and implicit coupling relationships embedded in mixed heterogeneous data from single/multiple sources. The coupling and heterogeneity of the non-IID aspects form the essence of big data and most real-world applications, namely the data is non-IID.

Most of classic theoretical systems and tools in statistics, data mining, database, knowledge management and machine learning assume the independence and identical distribution of underlying objects, features and values. Such theories and tools may lead to misleading or incorrect understanding of real-life data complexities. Non-IID learning in big data is a foundational theoretical problem in AI and data science, which considers the complex couplings and heterogeneity between entities, properties, interactions and contexts.

Topics of interest

Topics of interest include all aspects of learning from implicitly and/or explicitly non-IID data including, but not limited to:

Statistical foundation for non-IID learning
Mathematical foundation for non-IID learning
Probabilistic methods for non-IID learning
Statistical machine learning for non-IID learning
Non-IID learning theory and foundation
Non-IID data characterization
Non-IID data transformation
Non-IID data representation and encoding
Non-IID learning models and algorithms
Non-IID single-source analytics
Non-IID multi-source analytics
Non-IID clustering
Non-IID classification
Non-IID recommender systems
Non-IID text mining and document analysis
Non-IID image and video analytics

Organizers

Special Session Chairs:

Longbing Cao, Professor, Advanced Analytics Institute, University of Technology Sydney, Australia. Email: longbing.cao@uts.edu.au
Yang Gao, Professor, Department of Computer Science and Technology, Nanjing University, China. Email:gaoy@nju.edu.cn

Organization Chairs:

Yinghuan Shi, Department of Computer Science and Technology, Nanjing University, China. Email: syh@nju.edu.cn
Guansong Pang, Australian Institute of Machine Learning, University of Adelaide, Australia. Email: pangguansong@gmail.com
Chengzhang Zhu, Advanced Analytics Institute, University of Technology Sydney, Australia. Eamil: kevin.zhu.china@gmail.com

Important dates

Paper Submission: May 20, 2019
Notification of acceptance: July 25, 2019
Camera-Ready: Aug 8, 2019
Early Registration: Aug 15, 2019

About non-IID learning

While learning from non-IID data is not a new topic, which has been infrequently explored in various communities such as statistics, machine learning, data mining, and computer vision, we have seen much broader discussion on many other challenges and opportunities that have not been discussed previously, e.g., what is non-IID, the non-IIDness, relationship to dependence learning, statistical modeling, information theories, quantum computation, etc.

Readers of interest may refer to Non-IID Learning for more resources, information and discussion on non-IID learning about

What is non-IID learning and beyond IID?
The relevant research directions on non-IID learning
Relevant references on non-IID learning
Relevant activities on non-IID learning