Tune: A Research Platform for Distributed Model Selection and Training

Liaw, Richard; Liang, Eric; Nishihara, Robert; Moritz, Philipp; Gonzalez, Joseph E.; Stoica, Ion

doi:10.48550/arxiv.1807.05118

Cited by 240 publications

(283 citation statements)

References 7 publications

Supporting

Mentioning

243

Contrasting

Unclassified

Order By: Relevance

“…For network design and training, Tensorflow 2.0 (Abadi et al, 2016) and Keras 2.3 (Gulli & Pal, 2017) have been used. As model optimization framework, Tune 1.6.0 (Liaw et al, 2018) was selected.…”

Section: Appendixmentioning

confidence: 99%

A stacked deep convolutional neural network to predict the remaining useful life of a turbofan engine

Solis-Martin,

Galan-Paez,

Borrego-Diaz

2021

Preprint

View full text Add to dashboard Cite

This paper presents the data-driven techniques and methodologies used to predict the remaining useful life (RUL) of a fleet of aircraft engines that can suffer failures of diverse nature. The solution presented is based on two Deep Convolutional Neural Networks (DCNN) stacked in two levels. The first DCNN is used to extract a low-dimensional feature vector using the normalized raw data as input. The second DCNN ingests a list of vectors taken from the former DCNN and estimates the RUL. Model selection was carried out by means of Bayesian optimization using a repeated random subsampling validation approach. The proposed methodology was ranked in the third place of the 2021 PHM Conference Data Challenge.

show abstract

Section: Appendixmentioning

confidence: 99%

A stacked deep convolutional neural network to predict the remaining useful life of a turbofan engine

Solis-Martin,

Galan-Paez,

Borrego-Diaz

2021

Preprint

View full text Add to dashboard Cite

show abstract

“…A ResNet-50 [17] deep learning model, which was pre-trained on the ImageNet dataset [18], Model preparation and training were done using the PyTorch deep learning framework [19] and tuning of hyperparameters was done using the RayTune framework [20].…”

Section: Isolation Of Particulate Microscopy Imagesmentioning

confidence: 99%

Dataset of gold nanoparticle sizes and morphologies extracted from literature-mined microscopy images

Subramanian¹,

Cruse²,

Trewartha³

et al. 2021

Preprint

View full text Add to dashboard Cite

The factors controlling the size and morphology of nanoparticles have so far been poorly understood. Data-driven techniques are an exciting avenue to explore this field through the identification of trends and correlations in data. However, for these techniques to be utilized, large datasets annotated with the structural attributes of nanoparticles are required. While experimental SEM/TEM images collected from controlled experiments are reliable sources of this information, large-scale collection of these images across a variety of experimental conditions is expensive and infeasible.Published scientific literature, which provides a vast source of high-quality figures including SEM/TEM images, can provide a large amount of data at a lower cost if effectively mined. In this work, we develop an automated pipeline to retrieve and analyse microscopy images from gold nanoparticle literature and provide a dataset of 4361 SEM/TEM images of gold nanoparticles along with automatically extracted size and morphology information. The dataset can be queried to obtain information about the physical attributes of gold nanoparticles and their statistical distributions.

show abstract

“…We use the LBFGS optimizer (max iter=20, and history size=10) along with the Optuna library [3] in the Ray hyperparameter tuning framework [49]. Each dataset gets a budget of 200 trials to pick the best parameters on validation set.…”

Section: D1 Transfer Learningmentioning

confidence: 99%

Constrained Mean Shift Using Distant Yet Related Neighbors for Representation Learning

Tejankar¹,

Koohpayegani²,

Navaneet³

et al. 2021

Preprint

View full text Add to dashboard Cite

We are interested in representation learning in selfsupervised, supervised, or semi-supervised settings. The prior work on applying mean-shift idea for self-supervised learning, MSF, generalizes the BYOL idea by pulling a query image to not only be closer to its other augmentation, but also to the nearest neighbors (NNs) of its other augmentation. We believe the learning can benefit from choosing far away neighbors that are still semantically related to the query. Hence, we propose to generalize MSF algorithm by constraining the search space for nearest neighbors. We show that our method outperforms MSF in SSL setting when the constraint utilizes a different augmentation of an image, and outperforms PAWS in semi-supervised setting with less training resources when the constraint ensures the NNs have the same pseudolabel as the query.Our code is available here: https://github.com/UCDvision/CMSF.

show abstract

Tune: A Research Platform for Distributed Model Selection and Training

Cited by 240 publications

References 7 publications

A stacked deep convolutional neural network to predict the remaining useful life of a turbofan engine

A stacked deep convolutional neural network to predict the remaining useful life of a turbofan engine

Dataset of gold nanoparticle sizes and morphologies extracted from literature-mined microscopy images

Constrained Mean Shift Using Distant Yet Related Neighbors for Representation Learning

Contact Info

Product

Resources

About