Graph Kernel Neural Networks.

The convolution operator at the core of many modern neural architectures can effectively be seen as performing a dot product between an input matrix and a filter. While this is readily applicable to data such as images, which can be represented as regular grids in the Euclidean space, extending the convolution operator to work on graphs proves more challenging, due to their irregular structure. In this article, we propose to use graph kernels, i.e., kernel functions that compute an inner product on graphs, to extend the standard convolution operator to the graph domain. This allows us to define an entirely structural model that does not require computing the embedding of the input graph. Our architecture allows to plug-in any type of graph kernels and has the added benefit of providing some interpretability in terms of the structural masks that are learned during the training process, similar to what happens for convolutional masks in traditional convolutional neural networks (CNNs). We perform an extensive ablation study to investigate the model hyperparameters' impact and show that our model achieves competitive performance on standard graph classification and regression datasets.

Graph-in-Graph (GiG): Learning interpretable latent graphs in non-Euclidean domain for biological and healthcare applications.

Graphs are a powerful tool for representing and analyzing unstructured, non-Euclidean data ubiquitous in the healthcare domain. Two prominent examples are molecule property prediction and brain connectome analysis. Importantly, recent works have shown that considering relationships between input data samples has a positive regularizing effect on the downstream task in healthcare applications. These relationships are naturally modeled by a (possibly unknown) graph structure between input samples. In this work, we propose Graph-in-Graph (GiG), a neural network architecture for protein classification and brain imaging applications that exploits the graph representation of the input data samples and their latent relation. We assume an initially unknown latent-graph structure between graph-valued input data and propose to learn a parametric model for message passing within and across input graph samples, end-to-end along with the latent structure connecting the input graphs. Further, we introduce a Node Degree Distribution Loss (NDDL) that regularizes the predicted latent relationships structure. This regularization can significantly improve the downstream task. Moreover, the obtained latent graph can represent patient population models or networks of molecule clusters, providing a level of interpretability and knowledge discovery in the input domain, which is of particular value in healthcare.

Differentiable Graph Module (DGM) for Graph Convolutional Networks.

Graph deep learning has recently emerged as a powerful ML concept allowing to generalize successful deep neural architectures to non-euclidean structured data. Such methods have shown promising results on a broad spectrum of applications ranging from social science, biomedicine, and particle physics to computer vision, graphics, and chemistry. One of the limitations of the majority of current graph neural network architectures is that they are often restricted to the transductive setting and rely on the assumption that the underlying graph is known and fixed. Often, this assumption is not true since the graph may be noisy, or partially and even completely unknown. In such cases, it would be helpful to infer the graph directly from the data, especially in inductive settings where some nodes were not present in the graph at training time. Furthermore, learning a graph may become an end in itself, as the inferred structure may provide complementary insights next to the downstream task. In this paper, we introduce Differentiable Graph Module (DGM), a learnable function that predicts edge probabilities in the graph which are optimal for the downstream task. DGM can be combined with convolutional graph neural network layers and trained in an end-to-end fashion. We provide an extensive evaluation of applications from the domains of healthcare (disease prediction), brain imaging (age prediction), computer graphics (3D point cloud segmentation), and computer vision (zero-shot learning). We show that our model provides a significant improvement over baselines both in transductive and inductive settings and achieves state-of-the-art results.

Guided diffusion for inverse molecular design.

The holy grail of materials science is de novo molecular design, meaning engineering molecules with desired characteristics. The introduction of generative deep learning has greatly advanced efforts in this direction, yet molecular discovery remains challenging and often inefficient. Herein we introduce GaUDI, a guided diffusion model for inverse molecular design that combines an equivariant graph neural net for property prediction and a generative diffusion model. We demonstrate GaUDI's effectiveness in designing molecules for organic electronic applications by using single- and multiple-objective tasks applied to a generated dataset of 475,000 polycyclic aromatic systems. GaUDI shows improved conditional design, generating molecules with optimal properties and even going beyond the original distribution to suggest better molecules than those in the dataset. In addition to point-wise targets, GaUDI can also be guided toward open-ended targets (for example, a minimum or maximum) and in all cases achieves close to 100% validity of generated molecules.

An Information System for Brownfield Regeneration: providing customised information according to stakeholders' characteristics and needs.

In the EU brownfield presence is still considered a widespread problem. Even though, in the last decades, many research projects and initiatives developed a wealth of methods, guidelines, tools and technologies aimed at supporting brownfield regeneration. However, this variety of products had and still has a limited practical impact on brownfield revitalisation success, because they are not used in their entire potential due to their scarce visibility. Also, another problem that stakeholders face is finding customised information. To overcome this non-visibility and not-sufficient customisation of information, the Information System for Brownfield Regeneration (ISBR) has been developed, based on Artificial Neural Networks, which allows understanding stakeholders' information needs by providing tailored information. The ISBR has been tested by stakeholders from the EU project TIMBRE case studies, located in the Czech Republic, Germany, Poland and Romania. Data gained during tests allowed to understand stakeholders' information needs. Overall, stakeholders showed to be concerned first on remediation aspects, then on benchmarking information, which are valuable to improve practices in the complex field of brownfield regeneration, and then on the relatively new issue of sustainability applied to brownfield regeneration and remediation. Mature markets confirmed their interest for remediation-related aspects, highlighting the central role that risk assessment plays in the process. Emerging markets showed to seek information and tools for strategic and planning issues, like brownfield inventories and georeferenced data sets. Results led to conclude that a new improved platform, combining the ISBR functionalities with geo-referenced ones, would be useful and could represent a further research application.

An Accurate and Robust Artificial Marker Based on Cyclic Codes.

Artificial markers are successfully adopted to solve several vision tasks, ranging from tracking to calibration. While most designs share the same working principles, many specialized approaches exist to address specific application domains. Some are specially crafted to boost pose recovery accuracy. Others are made robust to occlusion or easy to detect with minimal computational resources. The sheer amount of approaches available in recent literature is indeed a statement to the fact that no silver bullet exists. Furthermore, this is also a hint to the level of scholarly interest that still characterizes this research topic. With this paper we try to add a novel option to the offer, by introducing a general purpose fiducial marker which exhibits many useful properties while being easy to implement and fast to detect. The key ideas underlying our approach are three. The first one is to exploit the projective invariance of conics to jointly find the marker and set a reading frame for it. Moreover, the tag identity is assessed by a redundant cyclic coded sequence implemented using the same circular features used for detection. Finally, the specific design and feature organization of the marker are well suited for several practical tasks, ranging from camera calibration to information payload delivery.