DeepGCNs: Can GCNs Go As Deep As CNNs?

Li, Guohao; Müller, Matthias; Thabet, Ali; Ghanem, Bernard

doi:10.1109/iccv.2019.00936

Cited by 1,045 publications

(684 citation statements)

References 30 publications

Supporting

Mentioning

608

Contrasting

Unclassified

Order By: Relevance

“…L2 regularization and dropout were found to be comparable to early-stopping on validation, so early-stop was used instead. Model training was found to diverge without residue connections, which others have seen [59] . Final layer numbers are K = 4, L = 3 , J = 3.…”

Section: Modelmentioning

confidence: 92%

Predicting Chemical Shifts with Graph Neural Networks

Yang

Chakraborty

White

2020

Preprint

View full text Add to dashboard Cite

Inferring molecular structure from NMR measurements requires an accurate forward model that can predict chemical shifts from 3D structure. Current forward models are limited to specific molecules like proteins and state of the art models are not differentiable. Thus they cannot be used with gradient methods like biased molecular dynamics. Here we use graph neural networks (GNNs) for NMR chemical shift prediction. Our GNN can model chemical shifts accurately and capture important phenomena like hydrogen bonding induced downfield shift between multiple proteins, secondary structure effects, and predict shifts of organic molecules. Previous empirical NMR models of protein NMR have relied on careful feature engineering with domain expertise. These GNNs are trained from data alone with no feature engineering yet are as accurate and can work on arbitrary molecular structures. The models are also efficient, able to compute one million chemical shifts in about 5 seconds. This work enables a new category of NMR models that have multiple interacting types of macromolecules.

show abstract

Section: Modelmentioning

confidence: 92%

Predicting Chemical Shifts with Graph Neural Networks

Yang

Chakraborty

White

2020

Preprint

View full text Add to dashboard Cite

show abstract

“…To capture higher-order dependencies, multiple dynamic graph convolution layers are chained and connected densely within a feature extraction unit [16,20,22,36]:…”

Section: Representation Encoder With Differentiable Poolingmentioning

confidence: 99%

Differentiable Manifold Reconstruction for Point Cloud Denoising

Luo

2020

Proceedings of the 28th ACM International Conference on Multimedia

View full text Add to dashboard Cite

3D point clouds are often perturbed by noise due to the inherent limitation of acquisition equipments, which obstructs downstream tasks such as surface reconstruction, rendering and so on. Previous works mostly infer the displacement of noisy points from the underlying surface, which however are not designated to recover the surface explicitly and may lead to sub-optimal denoising results. To this end, we propose to learn the underlying manifold of a noisy point cloud from differentiably subsampled points with trivial noise perturbation and their embedded neighborhood feature, aiming to capture intrinsic structures in point clouds. Specifically, we present an autoencoder-like neural network. The encoder learns both local and non-local feature representations of each point, and then samples points with low noise via an adaptive differentiable pooling operation. Afterwards, the decoder infers the underlying manifold by transforming each sampled point along with the embedded feature of its neighborhood to a local surface centered around the point. By resampling on the reconstructed manifold, we obtain a denoised point cloud. Further, we design an unsupervised training loss, so that our network can be trained in either an unsupervised or supervised fashion. Experiments show that our method significantly outperforms state-of-the-art denoising methods under both synthetic noise and real world noise. The code and data are available at https://github.com/luost26/DMRDenoise. CCS CONCEPTS • Computing methodologies → Point-based models; 3D imaging.

show abstract

“…Some of the key factors for using the Residual Network is the scalability of the training parameters by increasing the complexity of the network by adding additional layers. Like the classic Convolution Neural Network (CNNs), the Residual Network (ResNet) does not face the gradient issue of extinction [27].…”

Section: A Transfer Learningmentioning

confidence: 99%

Computer Vision and Radiology for COVID-19 Detection

Punia

Kumar

Mujahid

et al. 2020

2020 International Conference for Emerging Technology (INCET)

View full text Add to dashboard Cite

COVID-19 is spreading rapidly throughout the world. As of 14 April 2020, 128,000 people died of COVID-19, while 1.99 million cases in 210 countries and territories were reported in 219.747 cases. As the virus spreads at a very high rate, there is a huge shortage of medical testing kits all over the world. The respiratory system is the part of the human body most affected by the virus, so the use of X-rays of the chest may prove to be a more efficient way than the thermal screening of the human body. In this paper, we are trying to develop a method that uses radiology, i.e. X-rays for detecting the novel coronavirus. Along with the paper, we also release a dataset for the research community and further development extracted from various medical research hospital facilities treating COVID-19 patients.

show abstract

DeepGCNs: Can GCNs Go As Deep As CNNs?

Cited by 1,045 publications

References 30 publications

Predicting Chemical Shifts with Graph Neural Networks

Predicting Chemical Shifts with Graph Neural Networks

Differentiable Manifold Reconstruction for Point Cloud Denoising

Computer Vision and Radiology for COVID-19 Detection

Contact Info

Product

Resources

About