A5081978740- Advanced Neural Network Applications
- Machine Learning and Algorithms
- Ethics and Social Impacts of AI
- Domain Adaptation and Few-Shot Learning
- Scientific Measurement and Uncertainty Evaluation
- Machine Learning and Data Classification
- Brain Tumor Detection and Classification
- Autonomous Vehicle Technology and Safety
- Advanced Sensor Technologies Research
- Advanced Vision and Imaging
- Spectroscopy and Laser Applications
- Optical Network Technologies
- Calibration and Measurement Techniques
- Blockchain Technology Applications and Security
- Advanced Image and Video Retrieval Techniques
- Imbalanced Data Classification Techniques
- Robotics and Sensor-Based Localization
- Advanced Fiber Optic Sensors
- Advanced Optical Sensing Technologies
- Advanced Measurement and Metrology Techniques
- Advanced Fiber Laser Technologies
- Explainable Artificial Intelligence (XAI)
- Older Adults Driving Studies
- Flow Measurement and Analysis
- Transportation and Mobility Innovations
KU Leuven
2025
Scuola Normale Superiore
2022-2024
University of Pisa
2023
ALBA Synchrotron (Spain)
2023
Commonwealth Scientific and Industrial Research Organisation
2018
Centro de Investigación y Tecnología Agroalimentaria de Aragón
2008-2014
Universidad de Zaragoza
1984-2011
Institut d'Economie Scientifique Et de Gestion
1999
Humans can easily imagine the complete 3D geometry of occluded objects and scenes. This appealing ability is vital for recognition understanding. To enable such capability in AI systems, we propose VoxFormer, a Transformer-based semantic scene completion framework that output volumetric semantics from only 2D images. Our adopts two-stage design where start sparse set visible occupied voxel queries depth estimation, followed by densification stage generates dense voxels ones. A key idea this...
Abstract Turnover intention is an employee’s reported willingness to leave her organization within a given period of time and often used for studying actual employee turnover. Since turnover can have detrimental impact on business the labor market at large, it important understand determinants such choice. We describe analyze unique European-wide survey intention. A few baselines state-of-the-art classification models are compared as per predictive performances. Logistic regression LightGBM...
Abstract The literature addressing bias and fairness in AI models ( fair-AI ) is growing at a fast pace, making it difficult for novel researchers practitioners to have bird’s-eye view picture of the field. In particular, many policy initiatives, standards, best practices been proposed setting principles, procedures, knowledge bases guide operationalize management fairness. first objective this paper concisely survey state-of-the-art methods resources, main policies on AI, with aim providing...
Recent studies show that Vision Transformers(ViTs) exhibit strong robustness against various corruptions. Although this property is partly attributed to the self-attention mechanism, there still a lack of systematic understanding. In paper, we examine role in learning robust representations. Our study motivated by intriguing properties emerging visual grouping Transformers, which indicates may promote through improved mid-level We further propose family fully attentional networks (FANs)...
A mapping F2 population from the cross 'Piel de Sapo' × PI124112 was selectively genotyped to study genetic control of morphological fruit traits by QTL (Quantitative Trait Loci) analysis. Ten were identified, five for FL (Fruit Length), two FD Diameter) and three FS Shape). At least one robust per character found, flqs8.1 (LOD = 16.85, R2 34%), fdqs12.1 3.47, 11%) fsqs8.1 14.85, 41%). flqs2.1 fsqs2.1 cosegregate with gene a (andromonoecious), responsible flower sex determination pleiotropic...
We present counterfactual situation testing (CST), a causal data mining framework for detecting individual discrimination in dataset of classifier decisions. CST answers the question "what would have been model outcome had individual, or complainant, different protected status?" It extends legally-grounded (ST) Thanh et al. (2011) by operationalizing notion fairness given difference via reasoning. ST finds each complainant similar and non-protected instances dataset; constructs,...
Pruning aims to accelerate and compress models by removing redundant parameters, identified specifically designed importance scores which are usually imperfect. This removal is irreversible, often leading subpar performance in pruned models. Dynamic sparse training, while attempting adjust structures during training for continual reassessment refinement, has several limitations including criterion inconsistency between pruning growth, unsuitability structured sparsity, short-sighted growth...
As machine learning models enable decisions once performed only by humans, it is central to develop tools that assess the fairness of such models. Notably, within high-stake settings like hiring and lending, these must be able detect potentially discriminatory We present counterfactual situation testing (CST), a causal data mining framework for detecting individual discrimination in dataset classifier decisions. CST answers question “what would have been model outcome had individual, or...
Past research has demonstrated that the explicit use of protected attributes in machine learning can improve both performance and fairness. Many algorithms, however, cannot directly process categorical attributes, such as country birth or ethnicity. Because frequently are categorical, they must be encoded features input to a chosen algorithm, e.g. support vector machines, gradient boosting decision trees linear models. Thereby, encoding methods influence how what algorithm will learn,...
Structural pruning can simplify network architecture and improve inference speed. We propose Hardware-Aware Latency Pruning (HALP) that formulates structural as a global resource allocation optimization problem, aiming at maximizing the accuracy while constraining latency under predefined budget on targeting device. For filter importance ranking, HALP leverages lookup table to track reduction potential saliency score gauge drop. Both metrics be evaluated very efficiently during pruning,...
There is a fast-growing literature in addressing the fairness of AI models (fair-AI), with continuous stream new conceptual frameworks, methods, and tools. How much can we trust them? do they actually impact society? We take critical focus on fair-AI survey issues, simplifications, mistakes that researchers practitioners often underestimate, which turn undermine limit its contribution to society. In particular, discuss hyper-focus metrics optimizing their average performances. instantiate...
The advances in multimodal large language models (MLLMs) have led to growing interests LLM-based autonomous driving agents leverage their strong reasoning capabilities. However, capitalizing on MLLMs' capabilities for improved planning behavior is challenging since requires full 3D situational awareness beyond 2D reasoning. To address this challenge, our work proposes a holistic framework alignment between agent and tasks. Our starts with novel MLLM architecture that uses sparse queries lift...
In uses of pre-trained machine learning models, it is a known issue that the target population in which model being deployed may not have been reflected source with was trained. This can result biased when deployed, leading to reduction performance. One risk that, as changes, certain demographic groups will be under-served or otherwise disadvantaged by model, even they become more represented population. The field domain adaptation proposes techniques for situation where label data does...
Most current work on Boards of Directors has been focused what can be considered best practices for effective governance in terms roles, composition, process and style. Furthermore, this literature divided among very practically oriented managerial work, more rigorous, theoretically based work. the empirical research area centered Anglo‐Saxon countries. Some European countries, like Spain, have a different tradition. Given context, our present deals with two complementary objectives: 1. To...
We present counterfactual situation testing (CST), a causal data mining framework for detecting individual discrimination in dataset of classifier decisions. CST answers the question "what would have been model outcome had individual, or complainant, different protected status?" an actionable and meaningful way. It extends legally-grounded Thanh et al. [62] by operationalizing notion fairness given difference Kohler-Hausmann [38] using reasoning. In standard we find each complainant similar...
In this paper a technique consisting of measuring the Fourier transform ${Q}_{F}$ time-interval photon statistics distribution is studied when applied to laser Doppler velocimetry. It supposed that device changes Gaussian intensity profile beam into uniform used. A theoretical model for fluid with constant velocity obtained and verified by two ways: experimentally computer-simulation method. Then experimental conditions which signal can be approached Lorentzian curve error involved in...
Deep Neural Networks (DNNs) often rely on very large datasets for training. Given the size of such datasets, it is conceivable that they contain certain samples either do not contribute or negatively impact DNN's optimization. Modifying training distribution in a way excludes could provide an effective solution to both improve performance and reduce time. In this paper, we propose scale up ensemble Active Learning (AL) methods perform acquisition at (10k 500k time). We with ensembles...
In this paper we study the improvement in a laser Doppler velocimetry experiment when Fourier transform of time interval probability is measured instead intensity correlation function. The errors involved determination velocity are found to be greatly improved for low scattered intensities.
The common borage, Borago officinalis L., is of Euro-Mediterranean origin, but found extensively in the wild. It also cultivated as a garden plant, crop vegetable, or pharmaceutical herb. Analysis random amplified polymorphic DNAs (RAPDs) revealed high levels genetic diversity among 10 borage wild accessions and seven cultivars, which included five white-flowered selections grown northern Spain for petiole production, two blue-flowered lines used medicinally. These latter cultivars...