Adaptive Partial Scanning Transmission Electron Microscopy with\n Reinforcement Learning

Résumé fourni par la source

Compressed sensing can decrease scanning transmission electron microscopy\nelectron dose and scan time with minimal information loss. Traditionally,\nsparse scans used in compressed sensing sample a static set of probing\nlocations. However, dynamic scans that adapt to specimens are expected to be\nable to match or surpass the performance of static scans as static scans are a\nsubset of possible dynamic scans. Thus, we present a prototype for a contiguous\nsparse scan system that piecewise adapts scan paths to specimens as they are\nscanned. Sampling directions for scan segments are chosen by a recurrent neural\nnetwork based on previously observed scan segments. The recurrent neural\nnetwork is trained by reinforcement learning to cooperate with a feedforward\nconvolutional neural network that completes the sparse scans. This paper\npresents our learning policy, experiments, and example partial scans, and\ndiscusses future research directions. Source code, pretrained models, and\ntraining data is openly accessible at\nhttps://github.com/Jeffrey-Ede/adaptive-scans\n

Ce résumé expose les affirmations des auteurs. BNTIC ne l’interprète pas comme une validation indépendante des résultats.