arXiv subjects

Dominique Beaini

Publications and source records attributed to Dominique Beaini.

31 records · Page 2Linked to original sources

arXiv2023

Recipe for a General, Powerful, Scalable Graph Transformer

We propose a recipe on how to build a general, powerful, scalable (GPS) graph Transformer with linear complexity and state-of-the-art results on a diverse set of benchmarks. Graph Transformers (GTs) have gained popularity in the field of graph representation learning with a variety of recent publications but they lack a common foundation about what constitutes a good positional or structural encoding, and what differentiates them. In this paper, we summarize the different types of encodings with a clearer definition and categorize them as being $\textit{local}$, $\textit{global}$ or $\textit{relative}$. The prior GTs are constrained to small graphs with a few hundred nodes, here we propose the first architecture with a complexity linear in the number of nodes and edges $O(N+E)$ by decoupling the local real-edge aggregation from the fully-connected Transformer. We argue that this decoupling does not negatively affect the expressivity, with our architecture being a universal function approximator on graphs. Our GPS recipe consists of choosing 3 main ingredients: (i) positional/structural encoding, (ii) local message-passing mechanism, and (iii) global attention mechanism. We provide a modular framework $\textit{GraphGPS}$ that supports multiple types of encodings and that provides efficiency and scalability both in small and large graphs. We test our architecture on 16 benchmarks and show highly competitive results in all of them, show-casing the empirical benefits gained by the modularity and the combination of different strategies.

cs.LG↗

arXiv2022

GPS++: An Optimised Hybrid MPNN/Transformer for Molecular Property Prediction

This technical report presents GPS++, the first-place solution to the Open Graph Benchmark Large-Scale Challenge (OGB-LSC 2022) for the PCQM4Mv2 molecular property prediction task. Our approach implements several key principles from the prior literature. At its core our GPS++ method is a hybrid MPNN/Transformer model that incorporates 3D atom positions and an auxiliary denoising task. The effectiveness of GPS++ is demonstrated by achieving 0.0719 mean absolute error on the independent test-challenge PCQM4Mv2 split. Thanks to Graphcore IPU acceleration, GPS++ scales to deep architectures (16 layers), training at 3 minutes per epoch, and large ensemble (112 models), completing the final predictions in 1 hour 32 minutes, well under the 4 hour inference budget allocated. Our implementation is publicly available at: https://github.com/graphcore/ogb-lsc-pcqm4mv2.

q-bio.QM↗

arXiv2022

3D Infomax improves GNNs for Molecular Property Prediction

Molecular property prediction is one of the fastest-growing applications of deep learning with critical real-world impacts. Including 3D molecular structure as input to learned models improves their performance for many molecular tasks. However, this information is infeasible to compute at the scale required by several real-world applications. We propose pre-training a model to reason about the geometry of molecules given only their 2D molecular graphs. Using methods from self-supervised learning, we maximize the mutual information between 3D summary vectors and the representations of a Graph Neural Network (GNN) such that they contain latent 3D information. During fine-tuning on molecules with unknown geometry, the GNN still generates implicit 3D information and can use it to improve downstream tasks. We show that 3D pre-training provides significant improvements for a wide range of properties, such as a 22% average MAE reduction on eight quantum mechanical properties. Moreover, the learned representations can be effectively transferred between datasets in different molecular spaces.

cs.LG↗

arXiv2021

Rethinking Graph Transformers with Spectral Attention

In recent years, the Transformer architecture has proven to be very successful in sequence processing, but its application to other data structures, such as graphs, has remained limited due to the difficulty of properly defining positions. Here, we present the $\textit{Spectral Attention Network}$ (SAN), which uses a learned positional encoding (LPE) that can take advantage of the full Laplacian spectrum to learn the position of each node in a given graph. This LPE is then added to the node features of the graph and passed to a fully-connected Transformer. By leveraging the full spectrum of the Laplacian, our model is theoretically powerful in distinguishing graphs, and can better detect similar sub-structures from their resonance. Further, by fully connecting the graph, the Transformer does not suffer from over-squashing, an information bottleneck of most GNNs, and enables better modeling of physical phenomenons such as heat transfer and electric interaction. When tested empirically on a set of 4 standard datasets, our model performs on par or better than state-of-the-art GNNs, and outperforms any attention-based model by a wide margin, becoming the first fully-connected architecture to perform well on graph benchmarks.

cs.LG↗

arXiv2021

Directional Graph Networks

The lack of anisotropic kernels in graph neural networks (GNNs) strongly limits their expressiveness, contributing to well-known issues such as over-smoothing. To overcome this limitation, we propose the first globally consistent anisotropic kernels for GNNs, allowing for graph convolutions that are defined according to topologicaly-derived directional flows. First, by defining a vector field in the graph, we develop a method of applying directional derivatives and smoothing by projecting node-specific messages into the field. Then, we propose the use of the Laplacian eigenvectors as such vector field. We show that the method generalizes CNNs on an $n$-dimensional grid and is provably more discriminative than standard GNNs regarding the Weisfeiler-Lehman 1-WL test. We evaluate our method on different standard benchmarks and see a relative error reduction of 8% on the CIFAR10 graph dataset and 11% to 32% on the molecular ZINC dataset, and a relative increase in precision of 1.6% on the MolPCBA dataset. An important outcome of this work is that it enables graph networks to embed directions in an unsupervised way, thus allowing a better representation of the anisotropic features in different physical or biological problems.

cs.LG↗

arXiv2020

Principal Neighbourhood Aggregation for Graph Nets

Graph Neural Networks (GNNs) have been shown to be effective models for different predictive tasks on graph-structured data. Recent work on their expressive power has focused on isomorphism tasks and countable feature spaces. We extend this theoretical framework to include continuous features - which occur regularly in real-world input domains and within the hidden layers of GNNs - and we demonstrate the requirement for multiple aggregation functions in this context. Accordingly, we propose Principal Neighbourhood Aggregation (PNA), a novel architecture combining multiple aggregators with degree-scalers (which generalize the sum aggregator). Finally, we compare the capacity of different models to capture and exploit the graph structure via a novel benchmark containing multiple tasks taken from classical graph theory, alongside existing benchmarks from real-world domains, all of which demonstrate the strength of our model. With this work, we hope to steer some of the GNN research towards new aggregation methods which we believe are essential in the search for powerful and robust models.

cs.LG↗

arXiv2020

Towards Interpretable Sparse Graph Representation Learning with Laplacian Pooling

Recent work in graph neural networks (GNNs) has led to improvements in molecular activity and property prediction tasks. Unfortunately, GNNs often fail to capture the relative importance of interactions between molecular substructures, in part due to the absence of efficient intermediate pooling steps. To address these issues, we propose LaPool (Laplacian Pooling), a novel, data-driven, and interpretable hierarchical graph pooling method that takes into account both node features and graph structure to improve molecular representation. We benchmark LaPool on molecular graph prediction and understanding tasks and show that it outperforms recent GNNs. Interestingly, LaPool also remains competitive on non-molecular tasks. Both quantitative and qualitative assessments are done to demonstrate LaPool's improved interpretability and highlight its potential benefits in drug design. Finally, we demonstrate LaPool's utility for the generation of valid and novel molecules by incorporating it into an adversarial autoencoder.

cs.LG↗

arXiv2020

Improving Convolutional Neural Networks Via Conservative Field Regularisation and Integration

Current research in convolutional neural networks (CNN) focuses mainly on changing the architecture of the networks, optimizing the hyper-parameters and improving the gradient descent. However, most work use only 3 standard families of operations inside the CNN, the convolution, the activation function, and the pooling. In this work, we propose a new family of operations based on the Green's function of the Laplacian, which allows the network to solve the Laplacian, to integrate any vector field and to regularize the field by forcing it to be conservative. Hence, the Green's function (GF) is the first operation that regularizes the 2D or 3D feature space by forcing it to be conservative and physically interpretable, instead of regularizing the norm of the weights. Our results show that such regularization allows the network to learn faster, to have smoother training curves and to better generalize, without any additional parameter. The current manuscript presents early results, more work is required to benchmark the proposed method.

cs.CV↗

arXiv2020

Saliency Enhancement using Gradient Domain Edges Merging

In recent years, there has been a rapid progress in solving the binary problems in computer vision, such as edge detection which finds the boundaries of an image and salient object detection which finds the important object in an image. This progress happened thanks to the rise of deep-learning and convolutional neural networks (CNN) which allow to extract complex and abstract features. However, edge detection and saliency are still two different fields and do not interact together, although it is intuitive for a human to detect salient objects based on its boundaries. Those features are not well merged in a CNN because edges and surfaces do not intersect since one feature represents a region while the other represents boundaries between different regions. In the current work, the main objective is to develop a method to merge the edges with the saliency maps to improve the performance of the saliency. Hence, we developed the gradient-domain merging (GDM) which can be used to quickly combine the image-domain information of salient object detection with the gradient-domain information of the edge detection. This leads to our proposed saliency enhancement using edges (SEE) with an average improvement of the F-measure of at least 3.4 times higher on the DUT-OMRON dataset and 6.6 times higher on the ECSSD dataset, when compared to competing algorithm such as denseCRF and BGOF. The SEE algorithm is split into 2 parts, SEE-Pre for preprocessing and SEE-Post pour postprocessing.

cs.CV↗

arXiv2019

Deep Green Function Convolution for Improving Saliency in Convolutional Neural Networks

Current saliency methods require to learn large scale regional features using small convolutional kernels, which is not possible with a simple feed-forward network. Some methods solve this problem by using segmentation into superpixels while others downscale the image through the network and rescale it back to its original size. The objective of this paper is to show that saliency convolutional neural networks (CNN) can be improved by using a Green's function convolution (GFC) to extrapolate edges features into salient regions. The GFC acts as a gradient integrator, allowing to produce saliency features by filling thin edges directly inside the CNN. Hence, we propose the gradient integration and sum (GIS) layer that combines the edges features with the saliency features. Using the HED and DSS architecture, we demonstrated that adding a GIS layer near the network's output allows to reduce the sensitivity to the parameter initialization, to reduce the overfitting and to improve the repeatability of the training. By simply adding a GIS layer to the state-of-the-art DSS model, there is an absolute increase of 1.6% for the F-measure on the DUT-OMRON dataset, with only 10ms of additional computation time. The GIS layer further allows the network to perform significantly better in the case of highly noisy images or low-brightness images. In fact, we observed an F-measure improvement of 5.2% when noise was added to the dataset and 2.8% when the brightness was reduced. Since the GIS layer is model agnostic, it can be implemented into different fully convolutional networks. A major contribution of the current work is the first implementation of Green's function convolution inside a neural network, which allows the network to operate in the feature domain and in the gradient domain at the same time, thus improving the regional representation via edge filling.

cs.CV↗

arXiv2019

Computing the Spatial Probability of Inclusion inside Partial Contours for Computer Vision Applications

In Computer Vision, edge detection is one of the favored approaches for feature and object detection in images since it provides information about their objects boundaries. Other region-based approaches use probabilistic analysis such as clustering and Markov random fields, but those methods cannot be used to analyze edges and their interaction. In fact, only image segmentation can produce regions based on edges, but it requires thresholding by simply separating the regions into binary in-out information. Hence, there is currently a gap between edge-based and region-based algorithms, since edges cannot be used to study the properties of a region and vice versa. The objective of this paper is to present a novel spatial probability analysis that allows determining the probability of inclusion inside a set of partial contours (strokes). To answer this objective, we developed a new approach that uses electromagnetic convolutions and repulsion optimization to compute the required probabilities. Hence, it becomes possible to generate a continuous space of probability based only on the edge information, thus bridging the gap between the edge-based methods and the region-based methods. The developed method is consistent with the fundamental properties of inclusion probabilities and its results are validated by comparing an image with the probability-based estimation given by our algorithm. The method can also be generalized to take into consideration the intensity of the edges or to be used for 3D shapes. This is the first documented method that allows computing a space of probability based on interacting edges, which opens the path to broader applications such as image segmentation and contour completion.

cs.CV↗

arXiv2019

Fast and Optimal Laplacian Solver for Gradient-Domain Image Editing using Green Function Convolution

In computer vision, the gradient and Laplacian of an image are used in different applications, such as edge detection, feature extraction, and seamless image cloning. Computing the gradient of an image is straightforward since numerical derivatives are available in most computer vision toolboxes. However, the reverse problem is more difficult, since computing an image from its gradient requires to solve the Laplacian equation, also called Poisson equation. Current discrete methods are either slow or require heavy parallel computing. The objective of this paper is to present a novel fast and robust method of solving the image gradient or Laplacian with minimal error, which can be used for gradient domain editing. By using a single convolution based on a numerical Green's function, the whole process is faster and straightforward to implement with different computer vision libraries. It can also be optimized on a GPU using fast Fourier transforms and can easily be generalized for an n dimension image. The tests show that, for images of resolution 801x1200, the proposed GFC can solve 100 Laplacian in parallel in around 1.0 milliseconds ms. This is orders of magnitude faster than our nearest competitor which requires 294ms for a single image. Furthermore, we prove mathematically and demonstrate empirically that the proposed method is the least error solver for gradient domain editing. The developed method is also validated with examples of Poisson blending, gradient removal, and the proposed gradient domain merging GDM. Finally, we present how the GDM can be leveraged in future works for convolutional neural networks CNN.

cs.CV↗

arXiv2018

Novel Convolution Kernels for Computer Vision and Shape Analysis based on Electromagnetism

Computer vision is a growing field with a lot of new applications in automation and robotics, since it allows the analysis of images and shapes for the generation of numerical or analytical information. One of the most used method of information extraction is image filtering through convolution kernels, with each kernel specialized for specific applications. The objective of this paper is to present a novel convolution kernels, based on principles of electromagnetic potentials and fields, for a general use in computer vision and to demonstrate its usage for shape and stroke analysis. Such filtering possesses unique geometrical properties that can be interpreted using well understood physics theorems. Therefore, this paper focuses on the development of the electromagnetic kernels and on their application on images for shape and stroke analysis. It also presents several interesting features of electromagnetic kernels, such as resolution, size and orientation independence, robustness to noise and deformation, long distance stroke interaction and ability to work with 3D images

cs.CV↗