Abstract:In recent years, many spatial-temporal graph convolutional network (STGCN) models are proposed to deal with the spatial-temporal network data forecasting problem. These STGCN models have their own advantages, i.e., each of them puts forward many effective operations and achieves good prediction results in the real applications. If users can effectively utilize and combine these excellent operations integrating the advantages of existing models, then they may obtain more effective STGCN models thus create greater value using existing work. However, they fail to do so due to the lack of domain knowledge, and there is lack of automated system to help users to achieve this goal. In this article, we fill this gap and propose Auto-STGCN algorithm, which makes use of existing models to automatically explore high-performance STGCN model for specific scenarios. Specifically, we design Unified-STGCN framework, which summarizes the operations of existing architectures, and use parameters to control the usage and characteristic attributes of each operation, so as to realize the parameterized representation of the STGCN architecture and the reorganization and fusion of advantages. Then, we present Auto-STGCN, an optimization method based on reinforcement learning, to quickly search the parameter search space provided by Unified-STGCN, and generate optimal STGCN models automatically. Extensive experiments on real-world benchmark datasets show that our Auto-STGCN can find STGCN models superior to existing STGCN models used for search space construction, which demonstrates the effectiveness of our proposed method.

Autonomous Tool Construction with Gated Graph Neural Network

Mixed Geometry Message and Trainable Convolutional Attention Network for Knowledge Graph Completion

ToolNet: Using Commonsense Generalization for Predicting Tool Use for Robot Plan Synthesis

Visual Manipulation Relationship Detection based on Gated Graph Neural Network for Robotic Grasping

Robot-Enabled Construction Assembly with Automated Sequence Planning based on ChatGPT: RoboGPT

An Attentional Spatial Temporal Graph Convolutional Network with Co-Occurrence Feature Learning for Action Recognition

A stacked graph neural network with self-exciting process for robotic cognitive strategy reasoning in proactive human-robot collaborative assembly

Applying vision-guided graph neural networks for adaptive task planning in dynamic human robot collaborative scenarios

Auto-STGCN: Autonomous Spatial-Temporal Graph Convolutional Network Search

RPRG: Toward Real-time Robotic Perception, Reasoning and Grasping with One Multi-task Convolutional Neural Network.

Adversarial Object Rearrangement in Constrained Environments with Heterogeneous Graph Neural Networks

A Knowledge-Based Task Planning Approach for Robot Multi-Task Manipulation

Multi-Robot Active Mapping via Neural Bipartite Graph Matching

Self-supervised Graph Neural Network for Mechanical CAD Retrieval

Self-adaptive Multi-Robot Construction Using Gene Regulatory Networks

Auto-Learning-GCN: an Ingenious Framework for Skeleton-Based Action Recognition.

Geometric Task Networks: Learning Efficient and Explainable Skill Coordination for Object Manipulation

Automatic tower crane layout planning system for high-rise building construction using generative adversarial network

GNFactor: Multi-Task Real Robot Learning with Generalizable Neural Feature Fields

Learning Generalizable Tool-use Skills through Trajectory Generation

ToolNet: Connecting Large Language Models with Massive Tools via Tool Graph