Dr. Pavel Loskot, Zhejiang University-University of Illinois at Urbana-Champaign Institute (ZJUI)

Speech Abstract:

Monte Carlo methods are very attractive for numerically solving otherwise intractable mathematical problems such as computing integrals, solving differential equations, generating random processes, and analyzing and optimizing models of systems. In case of Monte Carlo simulations, the typical task is to find an empirical relationship between summary statistics of the model inputs and outputs. Whilst this approach is often deemed sufficient for validating engineering designs, its side effect is that even a complex model is then reduced into a mere transformation of model inputs to model outputs. This results in a substantial information loss as it hinders the internal system dynamics and workings. In order to overcome the information loss, the intermediate system outputs can be assumed to augment the global input-output summary statistics. In this talk, this problem will be discussed in the contexts of experiment design and sensitivity analysis for Monte Carlo simulations of engineering systems to allow extracting more knowledge about the model properties, and to improve the efficiency and statistical power of simulations by going beyond the input-output summary statistics.

Speech Abstract:

We are in the era of AI. Due to the extraordinary success of AI in many areas, the expectation from AI has skyrocketed. Should we be satisfied with the way things are going? It seems, implicitly we have started believing in philosophies like “bigger the better” (bigger data sets or bigger architecture with millions of free parameters) and “data say all”. And such approaches have been found to be very successful too. But is everything fine? Are we in the right direction? In this talk, I shall briefly mention the journey of AI to its present state and discuss some of the key limitations that demand attention. For example, usually most of the neural network (deep or shallow) based systems are neither comprehensible nor biologically plausible. Could these cause problems? Well, they at least raise some important concerns. In my view, the comprehensibility of a system depends, at least, on attributes like simplicity, transparency, explainability, trustworthiness, and in some cases the biological plausibility of such systems. Ideally, we should try to realize all these attributes in any AI system, but this is very difficult. I shall briefly discuss some of these issues where we need to pay more attention. I shall also allude to the role cognitive science can play in designing useful AI systems. Finally, I shall illustrate how one of those issues can be addressed taking inspiration from cat’s visual cortex.

Speech Abstract:

In order to improve the engineering feasibility of the memristive cellular neural network, a new memristor model with the smooth characteristic curve is designed. Based on the new memristor model, a new four-dimensional chaotic memristive cellular neural network system is constructed, and its chaotic dynamic behaviors are analyzed. Furthermore, in order to enhance the chaotic degree of cellular neural network (CNN), a five-dimensional memristive CNN hyperchaotic system is designed. Complex dynamic behaviors of memristive cellular neural network can be shown, and the circuit schematic diagrams of the systems can be designed. Improved sliding mode control method can be used to accomplish the chaos synchronization of memristive cellular neural network systems. Thus, chaotic memristive CNN system can be used in the secure communication by the chaos synchronization based on sliding mode control.

Speech Abstract:

The odor is the characteristic and alarming aroma of the human body. It is a significant information source of an individual's unique characteristic and physical condition in biometric, forensic and medical applications. Due to a complex combination of VOCs, the identification of individuals on the basis of body odor by conventional instruments is a tough task. The objective of the present research talk is to introduce audience about the data fusion and human body odor and to demonstrate research results related to search for an optimal subset of VOCs in body odor, which can produce differentiation in an individual by using the combination of analytical methods and chemometric analysis. Specifically, the implementation of data fusion approaches to search discriminating biomarker volatile organic chemicals (VOCs) in body odor for individual differentiation will be demonstrated. Also, some novel approaches to decision level data fusion will be discussed in human body odor mining. Gas chromatography–mass spectrometry (GC– MS) characterized human body odor samples have been used in analysis and validation of all experiments

Atabansi Chuwuemeka Clinton（Southwest University）

Feng Shao（Dalian Maritime University）