Reproduce “CityGPT: Towards Urban IoT Learning, Analysis and Interaction with Multi-Agent System” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Qinghua Guan, Jinhui Ouyang, Di Wu, Weiren Yu
CityGPT presents an end-to-end multi-agent framework, integrating Large Language Models and deep learning models to enhance urban IoT data analysis and interaction. The system enables non-expert users to derive actionable insights from complex spatio-temporal information through natural language queries, interactive visualizations, and descriptive text.
Key insights
- Integrating LLM-powered agents with deep learning analytics democratizes access to complex IoT data, enabling non-expert users to interact with and understand urban data through natural language.
- The use of a Parallel Genetic Algorithm (PGA) for hyperparameter optimization significantly improves the efficiency and deployment stability of deep learning models in dynamic IoT environments.
- Combining an Attentional GridLSTM with cross-attention for temporal dependencies and an extended CoSimHeat algorithm for spatial relationships effectively captures intricate spatio-temporal patterns in urban IoT data.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.