China
Africa
Explore chapters and articles related to this topic
Graph Representation Learning for Protein Classification
Published in Ranjeet Kumar Rout, Saiyed Umer, Sabha Sheikh, Amrit Lal Sangal, Artificial Intelligence Technologies for Computational Biology, 2023
Graph Neural Network (GNN) is an emerging branch of deep learning in non-Euclidean space, especially performing well in different tasks where graph-structured data are involved [19]. With the increasing growth of biological network data, the GNN approach has become an important tool in bioinformatics [4]. GNN approaches can extract structural and feature-related information from the graph and find euclidean representations for non-euclidean data, which help to perform several tasks like classification, regression, link prediction etc., on graphs [18]. GNNs have achieved excellent results in many biological tasks using different hierarchical views of a graph, which can include the collective behaviour of the nodes in the graph. Therefore, we believe that GNNs are potential enough to bridge the gap between the structure along with the function of a protein and the properties of the amino acids present in it.
Combining Theory and Data-Driven Approaches for Epidemic Forecasts
Published in Anuj Karpatne, Ramakrishnan Kannan, Vipin Kumar, Knowledge-Guided Machine Learning, 2023
Lijing Wang, Aniruddha Adiga, Jiangzhuo Chen, Bryan Lewis, Adam Sadilek, Srinivasan Venkatramanan, Madhav Marathe
Deep learning based methods have gained increasing prominence in epidemic forecasting due to their ability to learn non-linear relationship between the inputs and the outputs without prior domain knowledge. Some of the common structure of such networks include: feedforward neural networks (FNNs), recurrent neural networks (RNNs), convolutional neural networks (CNNs), and graph neural networks (GNNs). A FNN is an artificial neural network wherein connections between the nodes do not form a cycle. It was the first and simplest type of artificial neural network devised [58]. Forecasting prevalence of epidemics using FNNs is a widely accepted approach. For example, dengue forecasting by Wahyunggoro et al. [72] and Aburas et al. [1], and FNNs were first applied for influenza forecasting by Xu et al. [79]. Adhikari et al. [3] propose EpiDeep for seasonal ILI forecasting by learning meaningful representations of incidence curves in a continuous feature space. The RNNs, due to their ability to inherently capture the temporal dynamics, have become a natural choice for time series forecasting. Popular RNN modules are gated recurrent unit (GRU) [21] and long-short term memory (LSTM) [33]. Volkova et al. [71] build an LSTM model for short-term ILI forecasting using CDC ILI and Twitter data. Venna et al. [69] propose an LSTM-based method that integrates the impacts of climatic factors and geographical proximity. Zhu et al. [86] propose attention-based LSTM model for epidemic forecasting. Chimmula et al. [18] use LSTM networks to predict COVID-19 transmission. The CNNs are usually used to deal with image data with regular grid data structure. The idea is to sum the neighboring node features around a center node, specified by a filter with parameterized size and learnable weight. CNNs can be used for epidemic forecasting because multivariate time series (e.g., spatial regions) of an epidemic can be treated as an image with regular grid. Wu et al. [77] construct CNNRNN-Res combining RNN and convolutional neural networks to fuse information from different sources. The GNNs are the generalized version of CNN that can work on data with non-regular structures like a graph. The basic idea is to generate node embeddings based on local network neighborhoods through message passing. The neighborhoods are defined using an adjacency matrix. It can be any type of relationship between graph nodes. GNNs are famous for their ability to capture cross-spatial effects in dynamic environments, thus leading to an increased prominence in epidemic forecasting. Deng et al. [24] design cola-GNN, which is a cross-location attention-based graph neural network for forecasting ILI. Regarding COVID-19 forecasting, Kapoor et al. [38] and Wang et al. [74] examined graph neural networks for COVID-19 daily case prediction using mobility data. Aamchandani et al. [55] presented DeepCOVIDNet to compute equi-dimensional representations of multivariate time series.
Drainage pattern recognition method considering local basin shape based on graph neural network
Published in International Journal of Digital Earth, 2023
Wenning Wang, Haowen Yan, Xiaomin Lu, Yi He, Tao Liu, Wende Li, Pengbo Li, Fang Xu
Machine learning algorithms have been widely used in drug discovery (Dara et al. 2021), landslide prediction (He et al. 2021), and groundwater resources survey and assessment (Ruidas et al. 2021; Pal, Ruidas et al. 2022). However, vector data does not have a neat data arrangement structure, so it is difficult to use machine learning methods for vector data research. A graph neural network is a deep learning method based on a graph structure, and vector data can be transformed into graph structured data through certain transformations, thus graph deep learning is used for the study of vector data, which can effectively process and capture relational information in graphs by passing messages between graph nodes for tasks such as classification, prediction and clustering. The essence is extracting spatial features of topological graphs, mainly in the spatial and spectral domains, and based on this feature, to be widely used in vector data processing (Yu and Chen 2022; Yang, Yuan, et al. 2022).
Addressing cold start in recommender systems with neural networks: a literature survey
Published in International Journal of Computers and Applications, 2023
Graph Neural Networks (GNN): Graph neural networks are a class of NN that process data in the form of graphs. Article [21] uses pretraining GNN to simulate the cold start problem in order to improve predictions and adapt to real situations, while [29] uses a novel GNN that takes user preferences and attributes as well as items attributes and characteristics from an input layer and the last layer of the proposed framework is the prediction layer. Yet, another article [30] introduces a multilevel item hierarchy graph for items cold start, with a convolutional layer to decrease items parameters, whereas [39] shows that the trust network information and user ratings are aggregated by the GraphSAGE2 neural network algorithm to extract the user's hidden features vector.
AMGNET: multi-scale graph neural networks for flow field prediction
Published in Connection Science, 2022
Zhishuang Yang, Yidao Dong, Xiaogang Deng, Laiping Zhang
In recent years, there has been an increasing interest in applying graph neural networks to the task of processing graph-structured data. Graph neural networks (GNN) have been widely used in social networks, traffic speed prediction (Hu et al., 2022; Zhao et al., 2022), recommender systems and physics (Chang et al., 2021). Graph neural networks were first proposed by Gori et al. (2005) and Scarselli et al. (2009). Kipf and Welling (2016) proposed the graph convolutional operator and graph convolutional neural networks (GCN). Message passing neural networks unify various graph neural network and define the learning process of graph as Message Passing Phase and Readout Phase (Gilmer et al., 2017). Graph network (GN) proposed by Battaglia et al. (2018) is a flexible graph structure. Graph networks introduce inductive bias by constructing different aggregation and update functions.