Graph neural networks

Integrated World Modeling Theory (IWMT) is a synthetic theory of consciousness that uses the Free Energy Principle and Active Inference (FEP-AI) framework to combine insights from Integrated Information Theory (IIT) and Global Neuronal Workspace Theory (GNWT). Here, I first review philosophical principles and neural systems contributing to IWMT's integrative perspective. I then go on to describe predictive processing models of brains and their connections to machine learning architectures, with particular emphasis on autoencoders (perceptual and active inference), turbocodes (establishment of shared latent spaces for multi-modal integration and inferential synergy), and graph neural networks (spatial and somatic modeling and control). Particular emphasis is placed on the hippocampal/entorhinal system, which may provide a source of high-level reasoning via predictive contrasting and generalized navigation, so affording multiple kinds of conscious access. Future directions for IIT and GNWT are considered by exploring ways in which modules and workspaces may be evaluated as both complexes of integrated information and arenas for iterated Bayesian model selection. Based on these considerations, I suggest novel ways in which integrated information might be estimated using concepts from probabilistic graphical models, flow networks, and game theory. Mechanistic and computational principles are also considered with respect to the ongoing debate between IIT and GNWT regarding the physical substrates of different kinds of conscious and unconscious phenomena. I further explore how these ideas might relate to the "Bayesian blur problem", or how it is that a seemingly discrete experience can be generated from probabilistic modeling, with some consideration of analogies from quantum mechanics as potentially revealing different varieties of inferential dynamics. Finally, I go on to describe parallels between FEP-AI and theories of universal intelligence, including with respect to implications for the future of artificially intelligent systems. Particular emphasis is given to recurrent computation and its relationships with feedforward processing, including potential means of addressing critiques of causal structure theories based on network unfolding, and the seeming absurdity of conscious expander graphs (without cybernetic symbol grounding). While not quite solving the Hard problem, this article expands on IWMT as a unifying model of consciousness and the potential future evolution of minds.

This paper puts forth an aerial edge Internet-of-Things (EdgeIoT) system, where an unmanned aerial vehicle (UAV) is employed as a mobile edge server to process mission-critical computation tasks of ground Internet-of-Things (IoT) devices. When the UAV schedules an IoT device to offload its computation task, the tasks buffered at the other unselected devices could be outdated and have to be cancelled. We investigate a new joint optimization of UAV cruise control and task offloading allocation, which maximizes tasks offloaded to the UAV, subject to the IoT device’s computation capacity and battery budget, and the UAV’s speed limit. Since the optimization contains a large solution space while the instantaneous network states are unknown to the UAV, we propose a new deep graph-based reinforcement learning framework. An advantage actor-critic (A2C) structure is developed to train the real-time continuous actions of the UAV in terms of the flight speed, heading, and the offloading schedule of the IoT device. By exploring hidden representations resulting from the network feature correlation, our framework takes advantage of graph neural networks (GNN) to supervise the training of UAV’s actions in A2C. The proposed GNN-A2C framework is implemented with Google Tensorflow. The performance analysis shows that GNN-A2C achieves fast convergence and reduces considerably the task missing rate in aerial EdgeIoT.

Graph convolutional networks (GCNs) have received significant attention from various research fields due to the excellent performance in learning graph representations. Although GCN performs well compared with other methods, it still faces challenges. Training a GCN model for large-scale graphs in a conventional way requires high computation and storage costs. Therefore, motivated by an urgent need in terms of efficiency and scalability in training GCN, sampling methods have been proposed and achieved a significant effect. In this paper, we categorize sampling methods based on the sampling mechanisms and provide a comprehensive survey of sampling methods for efficient training of GCN. To highlight the characteristics and differences of sampling methods, we present a detailed comparison within each category and further give an overall comparative analysis for the sampling methods in all categories. Finally, we discuss some challenges and future research directions of the sampling methods.

Graph representation learning has received increasing attention in recent years for many real-world applications. A major challenge in graph representation learning is the lack of labeled data. To address this challenge, Graph Neural Networks (GNNs) use message passing frameworks to combine information from unlabeled data with labeled data. However, the use of unlabeled data under the message passing framework is indirect in the training process where unlabeled data does not supervise the training process. To fully exploit the potential of unlabeled data, we propose a novel dual-view cooperative training framework for graph data where unlabeled data is involved in the training process for supervision. Specifically, we regard different views as the reasoning processes of two GNN models with which the models make predictions, integrating the understanding of different models on the underlying graph. To exchange information between models, we design a pseudolabel-based approach, where the two models mutually provide pseudo labels to each other iteratively. Moreover, to ensure the quality of pseudo labels, we propose an entropy-based pseudo-labels selection procedure and we adopt GNNExplainer to visualize different views in our framework. Our comprehensive experimental evaluation shows that our methods can boost the performance of state-of-the-art models.

The task of inferring missing links or predicting future ones in a graph based on its current structure is referred to as link prediction. Link prediction methods that are based on pairwise node similarity are well-established approaches in the literature and show good prediction performance in many real-world graphs though they are heuristic. On the other hand, graph embedding approaches learn low-dimensional representation of nodes in graph and are capable of capturing inherent graph features, and thus support the subsequent link prediction task in graph. This appraisal paper studies a selection of methods from both categories on several benchmark (homogeneous) graphs with different properties from various domains. Beyond the intra and inter category comparison of the performances of the methods our aim is also to uncover interesting connections between Graph Neural Network(GNN)-based methods and heuristic ones as a means to alleviate the black-box well-known limitation.