Towards Augmented Microscopy with Reinforcement Learning-Enhanced Workflows. Xu, M., Kumar, A., & LeBeau, J. M Microsc. Microanal., 28(6):1–9, September, 2022. doi abstract bibtex Here, we report a case study implementation of reinforcement learning (RL) to automate operations in the scanning transmission electron microscopy workflow. To do so, we design a virtual, prototypical RL environment to test and develop a network to autonomously align the electron beam position without prior knowledge. Using this simulator, we evaluate the impact of environment design and algorithm hyperparameters on alignment accuracy and learning convergence, showing robust convergence across a wide hyperparameter space. Additionally, we deploy a successful model on the microscope to validate the approach and demonstrate the value of designing appropriate virtual environments. Consistent with simulated results, the on-microscope RL model achieves convergence to the goal alignment after minimal training. Overall, the results highlight that by taking advantage of RL, microscope operations can be automated without the need for extensive algorithm design, taking another step toward augmenting electron microscopy with machine learning methods.
@ARTICLE{Xu2022-fd,
title = "Towards Augmented Microscopy with Reinforcement
{Learning-Enhanced} Workflows",
author = "Xu, Michael and Kumar, Abinash and LeBeau, James M",
abstract = "Here, we report a case study implementation of reinforcement
learning (RL) to automate operations in the scanning transmission
electron microscopy workflow. To do so, we design a virtual,
prototypical RL environment to test and develop a network to
autonomously align the electron beam position without prior
knowledge. Using this simulator, we evaluate the impact of
environment design and algorithm hyperparameters on alignment
accuracy and learning convergence, showing robust convergence
across a wide hyperparameter space. Additionally, we deploy a
successful model on the microscope to validate the approach and
demonstrate the value of designing appropriate virtual
environments. Consistent with simulated results, the
on-microscope RL model achieves convergence to the goal alignment
after minimal training. Overall, the results highlight that by
taking advantage of RL, microscope operations can be automated
without the need for extensive algorithm design, taking another
step toward augmenting electron microscopy with machine learning
methods.",
journal = "Microsc. Microanal.",
volume = 28,
number = 6,
pages = "1--9",
month = sep,
year = 2022,
keywords = "automated microscopy; reinforcement learning; scanning
transmission electron microscopy;LeBeau Group;Amazon",
language = "en",
issn = "1431-9276, 1435-8115",
pmid = "36062363",
arxivid = "2208.02865",
doi = "10.1017/S1431927622012193"
}
Downloads: 0
{"_id":"8gTLfXzJTuN2EASCq","bibbaseid":"xu-kumar-lebeau-towardsaugmentedmicroscopywithreinforcementlearningenhancedworkflows-2022","author_short":["Xu, M.","Kumar, A.","LeBeau, J. M"],"bibdata":{"bibtype":"article","type":"article","title":"Towards Augmented Microscopy with Reinforcement Learning-Enhanced Workflows","author":[{"propositions":[],"lastnames":["Xu"],"firstnames":["Michael"],"suffixes":[]},{"propositions":[],"lastnames":["Kumar"],"firstnames":["Abinash"],"suffixes":[]},{"propositions":[],"lastnames":["LeBeau"],"firstnames":["James","M"],"suffixes":[]}],"abstract":"Here, we report a case study implementation of reinforcement learning (RL) to automate operations in the scanning transmission electron microscopy workflow. To do so, we design a virtual, prototypical RL environment to test and develop a network to autonomously align the electron beam position without prior knowledge. Using this simulator, we evaluate the impact of environment design and algorithm hyperparameters on alignment accuracy and learning convergence, showing robust convergence across a wide hyperparameter space. Additionally, we deploy a successful model on the microscope to validate the approach and demonstrate the value of designing appropriate virtual environments. Consistent with simulated results, the on-microscope RL model achieves convergence to the goal alignment after minimal training. Overall, the results highlight that by taking advantage of RL, microscope operations can be automated without the need for extensive algorithm design, taking another step toward augmenting electron microscopy with machine learning methods.","journal":"Microsc. Microanal.","volume":"28","number":"6","pages":"1–9","month":"September","year":"2022","keywords":"automated microscopy; reinforcement learning; scanning transmission electron microscopy;LeBeau Group;Amazon","language":"en","issn":"1431-9276, 1435-8115","pmid":"36062363","arxivid":"2208.02865","doi":"10.1017/S1431927622012193","bibtex":"@ARTICLE{Xu2022-fd,\n title = \"Towards Augmented Microscopy with Reinforcement\n {Learning-Enhanced} Workflows\",\n author = \"Xu, Michael and Kumar, Abinash and LeBeau, James M\",\n abstract = \"Here, we report a case study implementation of reinforcement\n learning (RL) to automate operations in the scanning transmission\n electron microscopy workflow. To do so, we design a virtual,\n prototypical RL environment to test and develop a network to\n autonomously align the electron beam position without prior\n knowledge. Using this simulator, we evaluate the impact of\n environment design and algorithm hyperparameters on alignment\n accuracy and learning convergence, showing robust convergence\n across a wide hyperparameter space. Additionally, we deploy a\n successful model on the microscope to validate the approach and\n demonstrate the value of designing appropriate virtual\n environments. Consistent with simulated results, the\n on-microscope RL model achieves convergence to the goal alignment\n after minimal training. Overall, the results highlight that by\n taking advantage of RL, microscope operations can be automated\n without the need for extensive algorithm design, taking another\n step toward augmenting electron microscopy with machine learning\n methods.\",\n journal = \"Microsc. Microanal.\",\n volume = 28,\n number = 6,\n pages = \"1--9\",\n month = sep,\n year = 2022,\n keywords = \"automated microscopy; reinforcement learning; scanning\n transmission electron microscopy;LeBeau Group;Amazon\",\n language = \"en\",\n issn = \"1431-9276, 1435-8115\",\n pmid = \"36062363\",\n arxivid = \"2208.02865\",\n doi = \"10.1017/S1431927622012193\"\n}\n\n","author_short":["Xu, M.","Kumar, A.","LeBeau, J. M"],"key":"Xu2022-fd","id":"Xu2022-fd","bibbaseid":"xu-kumar-lebeau-towardsaugmentedmicroscopywithreinforcementlearningenhancedworkflows-2022","role":"author","urls":{},"keyword":["automated microscopy; reinforcement learning; scanning transmission electron microscopy;LeBeau Group;Amazon"],"metadata":{"authorlinks":{}},"html":""},"bibtype":"article","biburl":"https://paperpile.com/eb/hvQdZzcQAp","dataSources":["T6bwdcdAx2jmtGv5a"],"keywords":["automated microscopy; reinforcement learning; scanning transmission electron microscopy;lebeau group;amazon"],"search_terms":["towards","augmented","microscopy","reinforcement","learning","enhanced","workflows","xu","kumar","lebeau"],"title":"Towards Augmented Microscopy with Reinforcement Learning-Enhanced Workflows","year":2022}