DIHGIGAL, Design and development of AI models, their training and validation

Design and development of generic models capable of adapting to learn to perform a specific task.

To this end, these models have a set of parameters that determine their behaviour and are adjusted to the appropriate value by means of a training process and on the basis of a set of data representative of the task to be learned.

Basically, in this process, the computer discovers how to solve the task according to the empirical data provided to it.

Design and development of generic models capable of adapting to learn to perform a specific task. To this end, these models have a set of parameters that determine their behaviour and are adjusted to the appropriate value by means of a training process and on the basis of a set of data representative of the task to be learned. Basically, in this process, the computer discovers how to solve the task according to the empirical data provided to it.

We can distinguish between supervised, unsupervised, semi-supervised and reinforcement learning techniques. The tasks that these algorithms can learn can also be stated, the most typical of which are classification, prediction, clustering, sorting and anomaly detection. Some applications include: Anomaly detection techniques Real-time and continuous learning Privacy-preserving distributed learning of data at individual sites.

The aim is to enable learning with data from different partners without compromising privacy between them.

• Learning techniques based on component analysis Evolutionary computing and bio-inspired or natural computation metaheuristics.
• Development of 'lightweight' models and algorithms for low-capacity devices and in connected environments (IoT).
• Feature engineering learning using perturbation theory in Bioinformatics/Chemoinformatics.
• Development of algorithms that take fairness issues into account.
• Explainability in machine learning.
• Machine learning with quantum computing.
• Human-in-the-loop machine learning.
• Machine learning for brain-computer interfaces.
• Behavioural learning and optimisation in mobile and autonomous robotics.
• Self-supervised and weakly supervised learning.
• Intelligent signal processing in mobile devices.
• Deep learning and new neural network architectures.
• Analysis of complex biological systems with machine learning and data integration techniques.
• Deterministic symbolic regression.
• Computational modelling.
• Deep learning for perception processing tasks, with focus on computational and environmental efficiency.
• Multi-task learning techniques for natural language Applications: video surveillance, image optimisation, music generation, signal analysis (especially medical signals), agriculture and food quality.