Konstantin Barkalov, Lipschitz global optimization and machine learning: helping each other to solve complex problems

Todor Ganchev, Intelligent Human-Machine Interaction (iHMI): Towards Adaptation to Momentary Emotional-Cognitive State of Users

Long Jin, Neural Dynamics Based Learning and  Control of Robots

Shuai Li , Beetle Antennae Search (BAS): Optimizing Unknown Functions Like a Beetle Searching for Food

Alexander Gornov, Computational technologies for approximating the reachable set of a nonlinear controlled system

Lipschitz global optimization and machine learning: helping each other to solve complex problems

In this talk we consider global optimization problems and methods for solving them. The numerical solution of this class of problems is fraught with significant difficulties. They are related to the dimensionality and type of the objective function. The most difficult problems are those in which the objective function is multi-extremal, nondifferentiable, and, moreover, given in the form of a “black box” (i.e., in the form of some computational procedure, the input of which is an argument, and the output is the corresponding value of the function). These complex problems (including multi-criteria problems) are the main focus of the talk. Particularly, we consider an approach to acceleration of the global search in multi-criteria problems using machine learning methods.

At the same time, the problem of tuning the hyperparameters of the machine learning methods themselves is very important. The quality of machine learning methods is substantially affected by their hyperparameters, while the evaluation of the quality metrics is a time-consuming operation. We also consider an approach to hyperparameter tuning based on the Lipschitz global optimization. These approaches are implemented in the iOpt open-source framework of intelligent optimization methods.

Intelligent Human-Machine Interaction (iHMI): Towards Adaptation to Momentary Emotional-Cognitive State of Users

The mass proliferation of AI technology and its deep penetration into the work and home routines of nearly everybody has become a game changer for modern society. In this regard, the availability of gentle human-centred interaction interfaces is among the essential aspects determining the user acceptability of AI-based technology, which might translate to their future social success or failure. This keynote discusses the limitations of current human-machine interfaces and the impact these restrictions have on their applications' usability and user acceptability. Furthermore, we present a unified framework that allows user-specific, task-related, and context-aware adaptations on the machine side to account for the user's momentary cognitive and emotional state. The practical significance of the proposed approach is demonstrated to allow for optimal performance for every person. The machine awareness of the individual optimal performance curves (Yerkes–Dodson law curves) permits the automated recognition of specific user profiles, real-time monitoring of the momentary condition, and activating a particular relationship management strategy. This is especially important when a change is detected caused by an alteration in the person’s state of mind under the influence of known or unknown factors.

Neural Dynamics Based Learning and  Control of Robots

The existing research has achieved success in learning and control of robots by using intelligent algorithms. However, most solutions require a large amount of data for network training, and there are some weaknesses, such as weak robustness and slow learning rates. To explore effective learning approaches, we establish a neural-dynamics-based unified framework for the learning and control of robots. Specifically, several synchronous learning and control strategies and corresponding neural dynamics are proposed for complex robot systems, such as multi-robot systems, robot visual servoing systems, mobile robot systems, and robot motion and force control systems. These research technologies include neural dynamics, data-driven technology, Kalman filtering algorithm, and robot technology. Abundant simulations, robotic experiments, and comparisons are carried out, and the results substantiate the robustness, practicability, and superiority of the proposed schemes when encountering uncertain situations.

Beetle Antennae Search (BAS): Optimizing Unknown Functions Like a Beetle Searching for Food

Effective optimization algorithms play a crucial role in real-world systems. Traditional gradient-based optimization methods have limitations in terms of the types of objective functions they can handle, as they often require prior knowledge of the system through an analytical model. Natureinspired metaheuristic optimization algorithms offer an efficient alternative to gradient-based techniques by emulating the behaviors of biological systems to tackle optimization challenges. This presentation centers around a significant algorithm that has recently garnered attention in research: the Beetle Antennae Search (BAS). BAS draws inspiration from the foraging habits of beetles and stands out for its unique approach. Unlike many swarm-based algorithms, BAS relies on single-search particles to seek optimal solutions, rendering it computationally efficient. BAS offers a noteworthy advantage in its ability to optimize system performance in a model-free manner, eliminating the need for an analytical system formula. Furthermore, its gradient-free nature allows it to excel in optimizing discrete systems as well. 

Computational technologies for approximating the reachable set of a nonlinear controlled system

The problem of phase estimation is considered, which belongs to the class of problems of constructing approximations of the reachable set for a controlled system. The controlled model is described by a system of nonlinear ordinary differential equations in the Cauchy normal form. Several algorithms for constructing an internal approximation of the reachable set based on the bang-bang principle and the cloud stochastic coverage method are discussed. The proposed computational technologies turn out to be a useful tool in creating algorithms for searching for a global extremum in optimal control problems. The results of computational experiments are presented.