In this paper, we elaborate on how AI can support diversity and inclusion exemplify research projects conducted in that direction. We start by looking at the challenges progress making large language models (LLMs) more transparent, inclusive, aware of social biases. Even though LLMs like ChatGPT have impressive abilities, they struggle to understand different cultural contexts engage meaningful, human conversations. A key issue is biases processing, especially machine translation, reinforce...

Accurate power load forecasting is essential for the efficient operation and planning of electrical grids, particularly given increased variability complexity introduced by renewable energy sources. This paper introduces GAT-LSTM, a hybrid model that combines Graph Attention Networks (GAT) Long Short-Term Memory (LSTM) networks. A key innovation incorporation edge attributes, such as line capacities efficiencies, into attention mechanism, enabling it to dynamically capture spatial...

Precisely estimating a child's body measurements and weight from single image is useful in pediatrics for monitoring growth detecting early signs of malnutrition. The development estimation models this task hampered by the unavailability labeled dataset to support supervised learning. This paper introduces "Age-Restricted Anonymized" (ARAN) dataset, first children with approved an ethics committee under European General Data Protection Regulation guidelines. ARAN consists images 512 aged 16...

Link prediction targets the of possible future links in a social network, i.e., we aim to predict next most likely network given current state. However, predicting solely based on (scarce) historic data is often challenging. In this paper, investigate, if can make use additional (domain) knowledge tackle problem. For purpose, apply answer set programming (ASP) for formalizing domain (and graph) analysis. particular, investigate link via ASP node proximity and its enhancement with background...

Abstract Using biased data to train Artificial Intelligence (AI) algorithms will lead decisions, discriminating against certain groups or individuals. Bias can be explicit (one several protected features directly influence the decisions) implicit indirectly decisions). Unsurprisingly, patterns are difficult detect and mitigate. This paper investigates extent which one more in structured classification sets mitigated simultaneously while retaining data’s discriminatory power. The main...

If C is a collection of mutually intersecting k-subsets fixed n-set, how big can be? And in the extreme case, what structure C? This question was answered by Erd?os, Ko, and Rado, (already 1938, but they first published this result 1961): One must have ... , if equality holds then all containing some element given n-set. The Kneser graph K(n, k) with as vertices where two are adjacent when disjoint. In terminology, Erdos, found largest cocliques (independent sets vertices) k). Many people...

In this study, we try to predict the Dutch housing market trends using text mining and machine learning as an application of data science methods in finance. Our main goal is short term upward or downward trend average house price by collected from Twitter. Twitter widely used well has been proven be a helpful source data. However, Twitter, (tokenization, bag-of-words, n-grams, weighted frequencies) (classification algorithms) have not combined yet order term. tweets including predefined...

Link prediction is challenging, especially based on (scarce) historic data or in cold start scenarios. In this paper, we show how to apply answer set programming (ASP) for formalizing link feature-rich networks, that - particular using domain knowledge network (and graph) analysis. We show, applying ASP provides a powerful declarative approach, as exemplified simple predictors, and demonstrate according explanation generation ASP. present the application of proposed methodological approach...

Malnutrition in children accounts for 45% of child deaths globally. Automatic malnutrition detection from digital photos serves as a decision support tool early rural areas. We study the feasibility estimating body-shape characteristics images human body s

For modeling user behavior in Artificial Intelligence (AI) systems, we can make use of diverse heterogeneous data sources. This paper investigates socio-spatial interaction networks for interactions from three perspectives: We analyze preferences and perceptions human social relation to the observed using wearable sensors. that, investigate correspondence according networks, order identify conformance, exceptions, anomalies. The analysis is performed on a real-world dataset capturing...

This paper investigates socio-spatial interaction networks for user modeling: We analyze preferences and perceptions of socio-proximity human interactions in relation to the observed interactions. The analysis is performed on a real-world dataset capturing using wearable sensors coupled with self-report questionnaires about perception those

Computational trust aims to quantify and is studied by many disciplines including computer science, social sciences business science. We propose a formal computational model, its parameters operations on these parameters, as well step guide compute in real application. make distinction between statements that aim capture the truth of dynamic phenomenon static phenomenon. elaborate how this difference should be reflected computation. To end we apply base rate (prior trust) an alternative...

The explication and the generation of explanations are prominent topics in artificial intelligence data science, order to make methods systems more transparent understandable for humans. This paper investigates problem link analysis, specifically prediction anomalous discovery social networks using declarative method Answer set programming (ASP). Applying ASP provides a powerful approach, e.g., incorporating domain knowledge explicative prediction. In this context, we propose novel...

We show that the Kneser graph of objects a fixed type in building spherical has unique coclique extension property when corresponding representation minuscule weight and also diagram is simply laced adjoint.

Consider the partial linear space on images in Λ/2Λ of shortest nonzero vectors Leech lattice Λ, where lines are triples adding up to zero. We determine universal embedding dimension and generating rank this (both 24) classify its hyperplanes.

This paper proposes an alternative approach for evidence based trust computation where the relationship between and is learned using artificial neuron, making it possible to automatically adapt different use cases. Computational aims quantify on ever increasing observations. In literature a value seen as posterior subjective probability, computed Bayesian inference evidence, prior weight of prior. provides fixed mapping trust, which may not be suitable every case study, e.g. when positive...