# Learning, Deciding & Acting in Complex Environments: Satellite Session

2017 Conference on Complex Systems

- Satellite Session planned for either September 20 & 21, 2017
- CSS venue Cancun, Mexico, September 17-22, 2016
- Session Organizers
- Kenric P. Nelson, Boston University, Boston, MA
- Mark Kon, Boston University, Boston, MA
- Sabir Umarov, University of New Haven, West Haven, CT
- Rudolf Hanel, Medical University of Vienna, Vienna, Austria

- Correspondence: kenricpn at bu dot edu, 781-645-8564

**Goal:** Develop a community of investigators applying the analytical methods of complex systems to the measurement, modeling, and design of systems which learn, decide and act upon information.

**Call for Papers:** Appropriate actions in a complex environment are driven by the capacity of a system to learn important features of signals in the environment and to be able to classify these signals into optimized decisions. Traditionally learning is a process of separating signals into deterministic components which are knowable from stochastic components which can be filtered. In complex environments an additional consideration is recognizing long-range patterns which are a mixture of deterministic and stochastic properties. Defining and measuring the complex components of signals, such as a nonlinear element which determines the rate of mixing between deterministic and stochastic elements, is an important element of designing improved algorithms. A characteristic example of this is a chaotic attractor which may originate from a simple nonlinear process but produces patterns with high entropy.

This CSS2017 session will focus on estimators and models that provide clear definitions of complexity which can be applied to improve decision making and enable control of actions to achieve specified objectives. Two analytical examples are the use of generalized maximum entropy methods to design learning algorithms, and the use of fractional calculus to design control filters. Two experimental examples of this are control of autonomous agents in a competitive situation and measuring the neural signals used by biological species to isolate the processes of decision-making and muscle control.

Papers are sought for the session which contribute analysis of rich datasets and/or improved mathematical models of complex environments. Submissions should show how improvements could be achieved in the understanding and application of one or more of the learning, deciding and acting processes. The authors should address how distinctions between deterministic, stochastic, and complex elements of signals are characterized. In particular how is complexity defined and how does this definition improve upon models which only address deterministic and stochastic elements.

**Themes:** Machine Learning, Decision & Information Theory, Control of Complex Systems, Fractional Calculus, Nonextensive Statistical Mechanics

**Submission by June 20, 2017:** https://easychair.org/conferences/?conf=ldace2017

**Image Citations:
**1. Balasis, G. et al. Statistical Mechanics and Information-Theoretic Perspectives on Complexity in the Earth System. Entropy 15, 4844–4888 (2013).

2. Motivated Machine Learning for Water Resource Management. at <https://www.slideshare.net/butest/motivated-machine-learning-for-water-resource-management>

*This webpage is maintained by the Boston University Department of Electrical and Computer Engineering.*