Towards generalizable deep ptychography neural networks

2025-09-29

AI Summary

Real-time feedback in X-ray ptychography is hindered by the poor generalizability and robustness of deep learning models across varying experimental conditions. To address this, we propose an unsupervised training framework centered on probe learning, which jointly incorporates experimentally measured probes and synthetically generated objects, embeds physics-informed neural networks, and introduces an out-of-distribution data augmentation strategy. Our method achieves, for the first time, generalized reconstruction across multiple beamlines and diverse probe types—delivering reconstruction accuracy comparable to that obtained using fully experimental training data, even under unseen experimental configurations. The model supports cross-facility deployment and enables millisecond-scale inference, fulfilling the stringent requirements of high-speed data acquisition and dynamic experiment control at next-generation synchrotron light sources. This work overcomes the critical robustness bottleneck imposed by experimental variability in existing ptychographic reconstruction approaches.

Technology Category

• Machine Learning: Calibration & Uncertainty Quantification
• Search and Optimization: Learning to Search
• Computer Vision: Learning & Optimization for CV

Application Category

• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Graph Algorithms and Modeling for the Web: Representation, reconstruction, and subgraph or motif discovery in Web-related graphs
• Security and Privacy: Data transparency and provenance

Abstract

X-ray ptychography is a data-intensive imaging technique expected to become ubiquitous at next-generation light sources delivering many-fold increases in coherent flux. The need for real-time feedback under accelerated acquisition rates motivates surrogate reconstruction models like deep neural networks, which offer orders-of-magnitude speedup over conventional methods. However, existing deep learning approaches lack robustness across diverse experimental conditions. We propose an unsupervised training workflow emphasizing probe learning by combining experimentally-measured probes with synthetic, procedurally generated objects. This probe-centric approach enables a single physics-informed neural network to reconstruct unseen experiments across multiple beamlines; among the first demonstrations of multi-probe generalization. We find probe learning is equally important as in-distribution learning; models trained using this synthetic workflow achieve reconstruction fidelity comparable to those trained exclusively on experimental data, even when changing the type of synthetic training object. The proposed approach enables training of experiment-steering models that provide real-time feedback under dynamic experimental conditions.

Problem

• Developing robust deep ptychography networks for diverse experimental conditions
• Achieving multi-probe generalization across different beamline configurations
• Enabling real-time reconstruction feedback under accelerated acquisition rates

Innovation

• Unsupervised training workflow with probe learning
• Combining experimental probes and synthetic objects
• Single physics-informed network for multi-probe generalization