Graph Neural Networks (GNNs) have exploded onto the machine learning scene in recent years owing to their capability to model and learn from graph-structured data. Such an ability has strong implications in a wide variety of fields whose data is inherently relational, for which conventional neural networks do not perform well. Indeed, as recent reviews can attest, research in the area of GNNs has grown rapidly and has lead to the development of a variety of GNN algorithm variants as well as to the exploration of groundbreaking applications in chemistry, neurology, electronics, or communication networks, among others. At the current stage of research, however, the efficient processing of GNNs is still an open challenge for several reasons. Besides of their novelty, GNNs are hard to compute due to their dependence on the input graph, their combination of dense and very sparse operations, or the need to scale to huge graphs in some applications. In this context, this paper aims to make two main contributions. On the one hand, a review of the field of GNNs is presented from the perspective of computing. This includes a brief tutorial on the GNN fundamentals, an overview of the evolution of the field in the last decade, and a summary of operations carried out in the multiple phases of different GNN algorithm variants. On the other hand, an in-depth analysis of current software and hardware acceleration schemes is provided, from which a hardware-software, graph-aware, and communication-centric vision for GNN accelerators is distilled.

The lack of fine-grained annotations hinders the deployment of automated diagnosis systems, which require human-interpretable justification for their decision process. In this paper, we address the problem of weakly supervised identification and localization of abnormalities in chest radiographs. To that end, we introduce a novel loss function for training convolutional neural networks increasing the \emph{localization confidence} and assisting the overall \emph{disease identification}. The loss leverages both image- and patch-level predictions to generate auxiliary supervision. Rather than forming strictly binary from the predictions as done in previous loss formulations, we create targets in a more customized manner, which allows the loss to account for possible misclassification. We show that the supervision provided within the proposed learning scheme leads to better performance and more precise predictions on prevalent datasets for multiple-instance learning as well as on the NIH~ChestX-Ray14 benchmark for disease recognition than previously used losses.

The task of writer verification is to provide a likelihood score for whether the queried and known handwritten image samples belong to the same writer or not. Such a task calls for the neural network to make it's outcome interpretable, i.e. provide a view into the network's decision making process. We implement and integrate cross-attention and soft-attention mechanisms to capture the highly correlated and salient points in feature space of 2D inputs. The attention maps serve as an explanation premise for the network's output likelihood score. The attention mechanism also allows the network to focus more on relevant areas of the input, thus improving the classification performance. Our proposed approach achieves a precision of 86\% for detecting intra-writer cases in CEDAR cursive "AND" dataset. Furthermore, we generate meaningful explanations for the provided decision by extracting attention maps from multiple levels of the network.

Can one learn to diagnose COVID-19 under extreme minimal supervision? Since the outbreak of the novel COVID-19 there has been a rush for developing Artificial Intelligence techniques for expert-level disease identification on Chest X-ray data. In particular, the use of deep supervised learning has become the go-to paradigm. However, the performance of such models is heavily dependent on the availability of a large and representative labelled dataset. The creation of which is a heavily expensive and time consuming task, and especially imposes a great challenge for a novel disease. Semi-supervised learning has shown the ability to match the incredible performance of supervised models whilst requiring a small fraction of the labelled examples. This makes the semi-supervised paradigm an attractive option for identifying COVID-19. In this work, we introduce a graph based deep semi-supervised framework for classifying COVID-19 from chest X-rays. Our framework introduces an optimisation model for graph diffusion that reinforces the natural relation among the tiny labelled set and the vast unlabelled data. We then connect the diffusion prediction output as pseudo-labels that are used in an iterative scheme in a deep net. We demonstrate, through our experiments, that our model is able to outperform the current leading supervised model with a tiny fraction of the labelled examples. Finally, we provide attention maps to accommodate the radiologist's mental model, better fitting their perceptual and cognitive abilities. These visualisation aims to assist the radiologist in judging whether the diagnostic is correct or not, and in consequence to accelerate the decision.

How can we efficiently compress a model while maintaining its performance? Knowledge Distillation (KD) is one of the widely known methods for model compression. In essence, KD trains a smaller student model based on a larger teacher model and tries to retain the teacher model's level of performance as much as possible. However, the existing KD methods suffer from the following limitations. First, since the student model is small in absolute size, it inherently lacks model complexity. Second, the absence of an initial guide for the student model makes it difficult for the student to imitate the teacher model to its fullest. Conventional KD methods yield low performance due to these limitations. In this paper, we propose Parameter-efficient and accurate Knowledge Distillation (Pea-KD), a novel approach to KD. Pea-KD consists of two main parts: Shuffled Parameter Sharing (SPS) and Pretraining with Teacher's Predictions (PTP). Using this combination, we are capable of alleviating the KD's limitations. SPS is a new parameter sharing method that allows greater model complexity for the student model. PTP is a KD-specialized initialization method, which can act as a good initial guide for the student. When combined, this method yields significant increase in student model's performance. Experiments conducted on different datasets and tasks show that the proposed approach improves the student model's performance by 4.4% on average in four GLUE tasks, outperforming existing KD baselines by significant margins.

We introduce Performers, Transformer architectures which can estimate regular (softmax) full-rank-attention Transformers with provable accuracy, but using only linear (as opposed to quadratic) space and time complexity, without relying on any priors such as sparsity or low-rankness. To approximate softmax attention-kernels, Performers use a novel Fast Attention Via positive Orthogonal Random features approach (FAVOR+), which may be of independent interest for scalable kernel methods. FAVOR+ can be also used to efficiently model kernelizable attention mechanisms beyond softmax. This representational power is crucial to accurately compare softmax with other kernels for the first time on large-scale tasks, beyond the reach of regular Transformers, and investigate optimal attention-kernels. Performers are linear architectures fully compatible with regular Transformers and with strong theoretical guarantees: unbiased or nearly-unbiased estimation of the attention matrix, uniform convergence and low estimation variance. We tested Performers on a rich set of tasks stretching from pixel-prediction through text models to protein sequence modeling. We demonstrate competitive results with other examined efficient sparse and dense attention methods, showcasing effectiveness of the novel attention-learning paradigm leveraged by Performers.

Effectively detecting anomalous nodes in attributed networks is crucial for the success of many real-world applications such as fraud and intrusion detection. Existing approaches have difficulties with three major issues: sparsity and nonlinearity capturing, residual modeling, and network smoothing. We propose Residual Graph Convolutional Network (ResGCN), an attention-based deep residual modeling approach that can tackle these issues: modeling the attributed networks with GCN allows to capture the sparsity and nonlinearity; utilizing a deep neural network allows to directly learn residual from the input, and a residual-based attention mechanism reduces the adverse effect from anomalous nodes and prevents over-smoothing. Extensive experiments on several real-world attributed networks demonstrate the effectiveness of ResGCN in detecting anomalies.

Competition between times series arises naturally in sales forecasting or in population modelisation. In this article, a model for the behavior of such high-dimensional time series is proposed. This model is based on the prerequisite that the total sum of the time series is distributed between its component following a 'competitiveness' factor inherent to each component. A confidence bound is proposed for the estimation of this model. Then, this model is applied to real-world E-commerce data using a recurrent neural network architecture to compute the model. It improves the results of standard RNN models in most cases.

In this paper, we discuss how modern deep learning approaches can be applied to the credit scoring of bank clients. We show that information about connections between clients based on money transfers between them allows us to significantly improve the quality of credit scoring compared to the approaches using information about the target client solely. As a final solution, we develop a new graph neural network model EWS-GCN that combines ideas of graph convolutional and recurrent neural networks via attention mechanism. The resulting model allows for robust training and efficient processing of large-scale data. We also demonstrate that our model outperforms the state-of-the-art graph neural networks achieving excellent results

Orthogonal weight matrices are used in many areas of deep learning. Much previous work attempt to alleviate the additional computational resources it requires to constrain weight matrices to be orthogonal. One popular approach utilizes *many* Householder reflections. The only practical drawback is that many reflections cause low GPU utilization. We mitigate this final drawback by proving that *one* reflection is sufficient, if the reflection is computed by an auxiliary neural network.