The following special sessions will be held during the conference. To submit a paper to one of these sessions, please complete the special session section of the paper submission form.
From Genetic Programming to Genetic Improvement Programming: Standing on the Shoulders of Giants
Genetic Programming is a supervised machine learning technique which constructs programs from scratch from a small instruction set. This has had its successes and produced human competitive results on a diverse number of domains including robot controllers, antenna design and a range of numerical algorithms. The algorithms evolved by genetic programming are typically small compared to typical engineered systems.
A complementary approach is to takes existing programs, engineered by humans, and apply the operators used by genetic programming (mutation and crossover). This is typically done by deleting, copying or exchanging lines of code within the program using the principles of natural evolution. This technique can be used to improve functional (or logical) properties of programs such as number of test cases passes (or an associated error score), or non-functional (or physical) properties such as runtime or power consumption. The resulting algorithm therefore is part manmade and part machine made. As genetic programming can be thought of as being inspired by natural evolution, genetic improvement programming can be thought of as being inspired by genetic modification.
Data Science and Heuristic Optimisation
Shahriar Asta and Ender Ozcan
Data science research is multidisciplinary with strong ties to the research in heuristic optimisation. While optimisation methods lie at the core of numerous data science techniques, heuristic optimisation methods increasingly use data science techniques to achieve improved performance or even generate new (components of) search methodologies. The aim of this stream is to explore the interaction between data science and heuristic optimisation. It will bring together researchers and practitioners from the fields of Operational Research, Artificial Intelligence, Mathematical Sciences and Computer Science, providing a medium for sharing and inspiring of new ideas and techniques, and developing common understandings.
Topics of interest include (but are not limited to):
- Improving heuristic optimisation using data science techniques, such as, machine learning, statistical learning, uncertainty modelling and more.
- (Meta)heuristics to optimise machine learning models.
- Machine learning for (meta)heuristic (component) generation.
- (Meta) heuristics for generating machine learning algorithms or their components.
- Achieving life-long learning via data science, (meta)heuristic optimisation or their hybrids.
- Systems teaching systems (e.g., apprenticeship learning, interactive evolutionary algorithms and inverse learning)
- (Meta)heuristics and data science for big data:
- Dimensionality reduction
- Pattern recognition
- Dynamic environments
- Streaming data
- Scalability issues
- A practical look at various machine learning and (meta)heuristic software libraries (e.g., Matlab, R, WEKA, HeuristicLab, Hyflex, Hyperion and more)
- Integration issues
- Parallelism/concurrency issues
- Extensions to existing libraries
- New problem instances
- Introducing new libraries
Big Data and Computational Intelligence in Networking (BDCI-NET 2015)
Yulei Wu and Geyong Min
Recent years have witnessed a deluge of network data propelled by the emerging online social media, user-generated video contents, and global-scale communications, bringing people into the era of big data. These network data hold much valuable information that could significantly improve the effective and intelligent optimisation of Internet. However, their unstructured, heterogeneous, sheer volume and complex nature of network big data pose great challenges on current computing, storage and networking architecture/protocols to automatically discover hidden information available for network optimisation. The existing state-of-the-art machine learning algorithms cannot scale up for big data analytics and large-scale networks due to very high computation overhead and non-real-time response. Therefore, it is crucial to devise efficient processing and analysis algorithms with significant computation intelligence to efficiently process the network big data.
This special session aims to address the key problems and challenges caused by network big data processing. The session will bring together computer scientists and engineers in different disciplines to share and exchange their experience and ideas, and discuss state-of-the-art and their in-progress research on big data and computation intelligence in networking.
Topics of particular interests include the following tracks, but are not limited to:
- Big Data Theory, Applications and Challenges for Networking
- Big Data Models and Algorithms for Networking
- Big Data Processing and Resource Scheduling for Networking
- Big Data for Network Protocol Design
- Big Data Computing, Storage and Networking Architectures
- Big Data Management and Distributed Data Systems
- Big Data Sharing, Security, Protection, Integrity, Trust and Privacy
- Big Data Mining and Analysis
- Big Data for Enterprise, Government and Society
- Mobility and Big Data
- Sensor Network, Social Network and Big Data
- Volume, Variety, Velocity, Value and Veracity (5V) of Big Data