Complex Cancer Detector: Complex Neural Networks on Non-stationary Time Series for Guiding Systematic Prostate Biopsy

Javadi, Golara; To, Minh Nguyen Nhat; Samadi, Samareh; Bayat, Sharareh; Sojoudi, Samira; Hurtado, Antonio; Chang, Silvia D.; Black, Peter C.; Mousavi, Parvin; Abolmaesumi, Purang

doi:10.1007/978-3-030-59716-0_50

Cited by 6 publications

(1 citation statement)

References 23 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…ML for non-stationary systems is an active field of research with recent results including the use of recurrent neural networks for speech perception in non-stationary noise [53], a method based on convolutional and recurrent neural networks has been proposed for modeling time-varying audio processors [54], complex-valued neural networks have been proposed for ML on non-stationary physical data and have shown that including phase information in feature maps improves both training and inference from deterministic physical data [55,56], and methods have been developed for non-stationary systems which detect significant changes after which the weights of neural networks are updated/re-trained with new information or are continuously trained to keep up with continuous changes [57][58][59]. Recently, a powerful class of approaches has been developed for the case of covariate shift, where the input distribution P (x) is different for training and test data, but the conditional distribution of output values P (y|x) remains unchanged [60], based on importance-weighting (IW) techniques [61].…”

Section: Non-stationary Distributionsmentioning

confidence: 99%

Adaptive Latent Space Tuning for Non-Stationary Distributions

Scheinker¹,

Cropp²,

Paiagua³

et al. 2021

Preprint

View full text Add to dashboard Cite

Powerful deep learning tools, such as convolutional neural networks (CNN), are able to learn the input-output relationships of large complicated systems directly from data. Encoder-decoder deep CNNs are able to extract features directly from images, mix them with scalar inputs within a general low-dimensional latent space, and then generate new complex 2D outputs which represent complex physical phenomenon. One important challenge faced by deep learning methods is large non-stationary systems whose characteristics change quickly with time for which re-training is not feasible. In this paper we present a method for adaptive tuning of the low-dimensional latent space of deep encoder-decoder style CNNs based on real-time feedback to quickly compensate for unknown and fast distribution shifts. We demonstrate our approach for predicting the properties of a time-varying charged particle beam in a particle accelerator whose components (accelerating electric fields and focusing magnetic fields) are also quickly changing with time.

show abstract