FutureAlgorithms - TalkAbstracts

Return to Conference Page

Prof. Tomas Recio

Title:
A real mechanical geometer

Abstract:
We introduce the popular dynamic geometry and computer algebra program GeoGebra (https://www.geogebra.org), with over 100 millions users all over the world, mostly in the educational context. Then we will exemplify the current performance of GeoGebra Discovery (https://github.com/kovzol/geogebra-discovery), a fork version of GeoGebra, that aims to be a sort of mechanical geometer, featuring automated reasoning tools for addressing elementary (but not trivial) geometry statements, by using computational algebraic geometry methods.

Finally, we focus on presenting some of GeoGebra Discovery pending algorithmic issues, such as implementing a cooperation with AI-bots, or extending the ability of GeoGebra Discovery to automatically discover new, interesting theorems.

(Joint work with Z. Kovács, Private University of Education, Diocese of Linz (Austria) and P. Vélez, U. Antonio de Nebrija, Madrid (Spain))

Prof. Jonathan Hauenstein

Title:
Computing smoothly connected components using routing functions

Abstract:
Nonlinear polynomial systems arise in many applications throughout science and engineering, and the physically-meaningful solutions typically arise from a subset of the real solution set. Some examples include an object's position in real affine space, while variables representing leg lengths or reaction rates are to be positive real numbers. This talk will explore an algorithm that uses routing functions to compute the smoothly connected components of the real solution set for a system of polynomial equations and inequalities. This algorithm will be illustrated on examples arising in science and engineering applications.

Dr Nikolas Papachristou

Title:
Palamedes: Algorithms and Architecture of a Superhuman AI for Backgammon and its Variants

Abstract:
This talk provides an in-depth exploration of Palamedes, a superhuman artificial intelligence system that excels at backgammon and its major variants. We present its end-to-end algorithmic framework, in which learning is driven by Temporal-Difference (TD) learning combined with neural networks for high-precision equity estimation. Superhuman performance is achieved through optimized N-ply search algorithms for move selection, complemented by specialized routines for match play—particularly addressing the strategic complexities of the three-game Tavli variant. From an engineering perspective, we highlight the modular system architecture that supports scalable training, competitive play, and advanced analysis across multiple platforms.

Dr Pengpeng Hu

Title:
Cost-effective 3D body scanning and measurement

Abstract:
3D body reconstruction is fundamental and important in various human-centric applications such as fashion design, animation, and healthcare. Traditional 3D scanners are accurate but costly and complex to operate, and often require users to remove clothing during data acquisition. The advent of commodity depth cameras enables affordable and accessible alternatives. Anthropometry, the scientific study of body measurement, offers a framework to capture human complexity and individuality. However, existing vision-based approaches still impose constraints on clothing and posture. To address these limitations, we aim at developing practical, comfortable, and privacy-aware solutions for 3D body reconstruction and human body measurement.

Dr Bahareh Afhami

Title:
Application of Principal Stratification in Crossover and Parallel Clinical Trial Designs

Abstract:
Estimating causal effects is crucial in clinical research for assessing treatment efficacy across various study designs. This paper explores causal effect estimation using the crossover and parallel models widely used in clinical trials. We investigate the application of principal stratification within both designs, focusing on key assumptions such as ignorability and no interference between units. Using a real-world example of basal insulin regimens, we compare treatment effect estimates derived from principal stratification with those obtained without these assumptions. Our findings illustrate that these methodologies yield reliable causal effect estimates, improving the accuracy of treatment effect evaluation in clinical trials.

Dr Abhishek Pal Majumdar

Title:

Abstract:

Dr Xiaolei Xie

Title:

Abstract:

Dr Omid Chatrabgoun

Title:

Abstract: