Local Construction

Local construction focuses on designing and optimizing algorithms and models that perform computations and inferences primarily using locally available data and resources, minimizing reliance on centralized servers or extensive communication. Current research emphasizes efficient local computation in diverse applications, including multi-agent systems (using architectures like team transformers), machine learning models (e.g., locally-editable GANs and retrieval-augmented LLMs), and privacy-preserving algorithms (like locally private contextual bandits). This approach is significant for improving efficiency, scalability, and privacy in various fields, ranging from software development and robotics to federated learning and statistical analysis.