Zero-Shot Forecasting of Network Dynamics through Weight Flow Matching

2025-10-09

AI Summary

Zero-shot prediction in networked systems (e.g., social media information diffusion) faces challenges due to persistent drifts in dynamic coefficients, scarcity of new labeled data, and prohibitions on retraining. Method: We propose a gradient-free cross-environment weight generation framework that jointly leverages a variational autoencoder (VAE) and conditional flow matching: the VAE encodes historical predictor weights into latent variables, while conditional flow matching learns a continuous, invertible mapping from a Gaussian prior to the target weight distribution under novel dynamic coefficients—enabling direct test-time generation of adapted model parameters. Contribution/Results: Our key innovation lies in formulating weight transfer as a conditional generative task, circumventing fine-tuning or retraining. Experiments across diverse, severe coefficient shift scenarios demonstrate substantial improvements over existing zero-shot baselines, with superior prediction stability and generalization.

Technology Category

• Machine Learning: Deep Generative Models & Autoencoders
• Computer Vision: Generative Adversarial Networks (GANs) for Vision
• Natural Language Processing: Generation

Application Category

• Social Networks and Social Media: Generative AI / large language models and their impact on social systems
• Graph Algorithms and Modeling for the Web: Foundation models and LLMs for Web-related graphs
• Web Mining and Content Analysis: Large pretrained models with web data

Abstract

Forecasting state evolution of network systems, such as the spread of information on social networks, is significant for effective policy interventions and resource management. However, the underlying propagation dynamics constantly shift with new topics or events, which are modeled as changing coefficients of the underlying dynamics. Deep learning models struggle to adapt to these out-of-distribution shifts without extensive new data and retraining. To address this, we present Zero-Shot Forecasting of Network Dynamics through Weight Flow Matching (FNFM), a generative, coefficient-conditioned framework that generates dynamic model weights for an unseen target coefficient, enabling zero-shot forecasting. Our framework utilizes a Variational Encoder to summarize the forecaster weights trained in observed environments into compact latent tokens. A Conditional Flow Matching (CFM) module then learns a continuous transport from a simple Gaussian distribution to the empirical distribution of these weights, conditioned on the dynamical coefficients. This process is instantaneous at test time and requires no gradient-based optimization. Across varied dynamical coefficients, empirical results indicate that FNFM yields more reliable zero-shot accuracy than baseline methods, particularly under pronounced coefficient shift.

Problem

• Forecasting network state evolution under shifting dynamics
• Adapting to out-of-distribution shifts without retraining
• Generating dynamic model weights for unseen target coefficients

Innovation

• Generative framework creates dynamic model weights
• Variational Encoder compresses weights into latent tokens
• Conditional Flow Matching enables zero-shot forecasting