Deep Learning

Deep learning is a subset of machine learning based on artificial neural networks with multiple layers to extract higher-level features from data.

Deep learning utilizes multi-layered artificial neural networks to model and solve complex problems. By using deep architectures, these models can automatically learn representations from raw data without the need for manual feature engineering, making them highly effective for processing high-dimensional and non-linear data.

The multi-layered architecture of a typical deep neural network.

Our research focuses on applying deep learning to the domain of communication systems, investigating architectures such as Graph Neural Networks (GNNs) and Transformers to handle structured and relational data in networks. These models enable more efficient and accurate performance prediction and system modeling by capturing complex spatial and temporal dependencies.

We also explore the challenges of training deep models for real-time applications, including computational efficiency and generalization. Our work aims to develop robust architectures that can operate reliably even in data-constrained environments or when faced with highly dynamic network conditions.