scholarly journals AutoReCon: Neural Architecture Search-based Reconstruction for Data-free Compression

2021 ◽  
Author(s):  
Baozhou Zhu ◽  
Peter Hofstee ◽  
Johan Peltenburg ◽  
Jinho Lee ◽  
Zaid Alars
Keyword(s):  
Experimental Results ◽  
Training Dataset ◽  
Reconstruction Method ◽  
Original Training ◽  
Neural Architecture ◽  
Reconstruction Methods ◽  
Network Engineering ◽  
Current Reconstruction

Data-free compression raises a new challenge because the original training dataset for a pre-trained model to be compressed is not available due to privacy or transmission issues. Thus, a common approach is to compute a reconstructed training dataset before compression. The current reconstruction methods compute the reconstructed training dataset with a generator by exploiting information from the pre-trained model. However, current reconstruction methods focus on extracting more information from the pre-trained model but do not leverage network engineering. This work is the first to consider network engineering as an approach to design the reconstruction method. Specifically, we propose the AutoReCon method, which is a neural architecture search-based reconstruction method. In the proposed AutoReCon method, the generator architecture is designed automatically given the pre-trained model for reconstruction. Experimental results show that using generators discovered by the AutoRecon method always improve the performance of data-free compression.

scholarly journals A Current Reconstruction Strategy Following the Operation Area in a 1-Shunt Inverter System

Energies ◽  
2019 ◽  
Vol 12 (8) ◽  
pp. 1423 ◽  
Author(s):  
Chang-Hwan Park ◽  
Dong-Youn Kim ◽  
Han-Beom Yeom ◽  
Yung-Deug Son ◽  
Jang-Mok Kim
Keyword(s):  
Acoustic Noise ◽  
Harmonic Distortion ◽  
Reconstruction Method ◽  
Reconstruction Methods ◽  
Modulation Region ◽  
Average Current ◽  
Conventional Methods ◽  
The Cost ◽  
Current Reconstruction ◽  
A Current

To reduce the cost of the inverter system in home appliances, a method using a shunt resistor at the DC-link can be substituted for a method using two current sensors at the inverter output. However, the minimum time of the active vector is required to sample the accurate current using a 1-shunt resistor. Therefore, many studies have been conducted investigating current reconstruction methods in the unmeasurable region of the current. The conventional methods using voltage injection have some problems such as high THD (Total Harmonic Distortion) and acoustic noise, because the PWM pattern is shifted. In addition, the current reconstruction is inaccurate in a low modulation region. In this paper, the cause of the noise in conventional methods is analyzed and a simple current reconstruction method based on an average current estimation and a current reference is utilized for reducing acoustic noise. In an immeasurable area, especially a low modulation region, an intermittent PWM shift method is proposed to enhance the accuracy of the reconstructed current. Therefore, a control strategy that combines all of the mentioned methods is implemented for the entire operating range. The effectiveness of the proposed methods is verified through the experimental results and the results of sound measurement in an anechoic chamber are included to compare with the acoustic noise.

scholarly journals A Novel Phase Current Reconstruction Method for a Three-Level Neutral Point Clamped Inverter (NPCI) with a Neutral Shunt Resistor

Energies ◽  
2018 ◽  
Vol 11 (10) ◽  
pp. 2616 ◽  
Author(s):  
Yungdeug Son ◽  
Jangmok Kim
Keyword(s):  
Estimation Method ◽  
Harmonic Distortion ◽  
Current Measurement ◽  
Neutral Point ◽  
Minimum Time ◽  
Reconstruction Method ◽  
Phase Current ◽  
Reconstruction Methods ◽  
Restoration Method ◽  
Current Reconstruction

This paper presents three phase current reconstruction methods for a three-level neutral point clamped inverter (NPCI) by measuring the voltage of a shunt resistor placed in the neutral point of the inverter. In order to accurately acquire the phase currents from the shunt resister, the dwell time of the active voltage vectors need to exceed the minimum time. On the other hand, if the time of active voltage is shorter than the minimum time, the current measurement becomes impossible. In this paper, unmeasurable regions for current are classified into three areas. Area 1 is a region in which both phase currents can be measure. Therefore, it is not necessary to restore the current. In Area 2, only one phase current can be measured. Thus, an estimation or restoration method is needed to measure another phase current. In this paper, the current estimation method using an electrical model of the motor is proposed. Area 3 is the region in which both phase currents can not be measured. In this case, it is necessary to move the voltage vector to the current measurable area by injecting the voltage. In this paper, Area 3 is divided into 36 sectors to inject optimal voltage. The proposed methods have the advantages of high current measurement accuracy and low THD (total harmonic distortion). The effectiveness of the proposed methods are verified through experimental results.

Global Inhomogeneity Index Evaluation of a DCT-based EIT Lung Imaging

2020 ◽  
Vol 6 (3) ◽  
pp. 36-39
Author(s):  
Rongqing Chen ◽  
Knut Möller
Keyword(s):  
Classical Method ◽  
Lung Imaging ◽  
Imaging Method ◽  
Reconstruction Method ◽  
Essential Information ◽  
Preliminary Results ◽  
Reconstruction Methods ◽  
Comparison Results ◽  
Index Evaluation ◽  
Inhomogeneity Index

AbstractPurpose: To evaluate a novel structural-functional DCT-based EIT lung imaging method against the classical EIT reconstruction. Method: Taken retrospectively from a former study, EIT data was evaluated using both reconstruction methods. For different phases of ventilation, EIT images are analyzed with respect to the global inhomogeneity (GI) index for comparison. Results: A significant less variant GI index was observed in the DCTbased method, compared to the index from classical method. Conclusion: The DCT-based method generates more accurate lung contour yet decreasing the essential information in the image which affects the GI index. These preliminary results must be consolidated with more patient data in different breathing states.

scholarly journals Deep Learning-Based Computer-Aided Diagnosis System for Gastroscopy Image Classification Using Synthetic Data

2021 ◽  
Vol 11 (2) ◽  
pp. 760
Author(s):  
Yun-ji Kim ◽  
Hyun Chin Cho ◽  
Hyun-chong Cho
Keyword(s):  
Deep Learning ◽  
Disease Diagnosis ◽  
Computer Aided Diagnosis ◽  
Medical Data ◽  
Classification Model ◽  
Training Dataset ◽  
Generative Adversarial Networks ◽  
Original Training ◽  
Computer Aided ◽  
Aided Diagnosis

Gastric cancer has a high mortality rate worldwide, but it can be prevented with early detection through regular gastroscopy. Herein, we propose a deep learning-based computer-aided diagnosis (CADx) system applying data augmentation to help doctors classify gastroscopy images as normal or abnormal. To improve the performance of deep learning, a large amount of training data are required. However, the collection of medical data, owing to their nature, is highly expensive and time consuming. Therefore, data were generated through deep convolutional generative adversarial networks (DCGAN), and 25 augmentation policies optimized for the CIFAR-10 dataset were implemented through AutoAugment to augment the data. Accordingly, a gastroscopy image was augmented, only high-quality images were selected through an image quality-measurement method, and gastroscopy images were classified as normal or abnormal through the Xception network. We compared the performances of the original training dataset, which did not improve, the dataset generated through the DCGAN, the dataset augmented through the augmentation policies of CIFAR-10, and the dataset combining the two methods. The dataset combining the two methods delivered the best performance in terms of accuracy (0.851) and achieved an improvement of 0.06 over the original training dataset. We confirmed that augmenting data through the DCGAN and CIFAR-10 augmentation policies is most suitable for the classification model for normal and abnormal gastric endoscopy images. The proposed method not only solves the medical-data problem but also improves the accuracy of gastric disease diagnosis.

scholarly journals Novel Reconstruction Using Roux-en-Y Plus Antral Obstruction Following Proximal Gastrectomy: Feasibility and Quality of Life Benefits

2019 ◽  
Author(s):  
Dejun Yang ◽  
Changming Wang ◽  
Hongbing Fu ◽  
Ziran Wei ◽  
Xin Zhang ◽  
...  
Keyword(s):  
Quality Of Life ◽  
Reflux Esophagitis ◽  
Proximal Gastrectomy ◽  
Partial Gastrectomy ◽  
Incidence Rates ◽  
Reconstruction Method ◽  
Reflux Symptoms ◽  
Reconstruction Methods ◽  
Postoperative Reflux

Abstract Background and Aims Routine gastroesophagostomy has been shown to have adverse effects on the recovery of digestive functions and quality of life because patients typically experience reflux symptoms after proximal gastrectomy. This study was performed to assess the feasibility and quality of life benefits of a novel reconstruction method termed Roux-en-Y anastomosis plus antral obstruction (RYAO) following proximal partial gastrectomy. Methods A total of 73 patients who underwent proximal gastrectomy from June 2015 to June 2017 were divided into two groups according to digestive reconstruction methods [RYAO (37 patients) and conventional esophagogastric anastomosis with pyloroplasty (EGPP, 36 patients)]. Clinical data were compared between the two groups retrospectively. Results The mean operative time for digestive reconstruction was slightly longer in the RYAO group than in the EGPP group. However, the incidence of postoperative short-term complications did not differ between the RYAO and the EGPP groups. At the 6-month follow-up, the incidence rates of both reflux esophagitis and gastritis were lower in the RYAO group than in the EGPP group (P = 0.002). Additionally, body weight recovery was better in the RYAO group (P = 0.028). The scale tests indicated that compared with the patients in the EGPP group, the patients in the RYAO group had significantly reduced reflux, nausea and vomiting and reported improvements in their overall health status and quality of life (all P < 0.05). Conclusion RYAO reconstruction may be a feasible procedure to reduce postoperative reflux symptoms and the incidence of reflux esophagitis and gastritis, thus improving patient quality of life after proximal gastrectomy.

scholarly journals Reconstructing intragranular strain fields in polycrystalline materials from scanning 3DXRD data

2020 ◽  
Vol 53 (2) ◽  
pp. 314-325 ◽  
Author(s):  
N. Axel Henningsson ◽  
Stephen A. Hall ◽  
Jonathan P. Wright ◽  
Johan Hektor
Keyword(s):  
Three Dimensional ◽  
Polycrystalline Materials ◽  
Reconstruction Method ◽  
X Ray Diffraction ◽  
Reconstruction Accuracy ◽  
Strain Fields ◽  
Spatial Properties ◽  
Reconstruction Methods ◽  
Reconstruction Quality ◽  
Made In

Two methods for reconstructing intragranular strain fields are developed for scanning three-dimensional X-ray diffraction (3DXRD). The methods are compared with a third approach where voxels are reconstructed independently of their neighbours [Hayashi, Setoyama & Seno (2017). Mater. Sci. Forum, 905, 157–164]. The 3D strain field of a tin grain, located within a sample of approximately 70 grains, is analysed and compared across reconstruction methods. Implicit assumptions of sub-problem independence, made in the independent voxel reconstruction method, are demonstrated to introduce bias and reduce reconstruction accuracy. It is verified that the two proposed methods remedy these problems by taking the spatial properties of the inverse problem into account. Improvements in reconstruction quality achieved by the two proposed methods are further supported by reconstructions using synthetic diffraction data.

Efficient 3D elastic full-waveform inversion using wavefield reconstruction methods

Geophysics ◽  
2016 ◽  
Vol 81 (2) ◽  
pp. R45-R55 ◽  
Author(s):  
Espen Birger Raknes ◽  
Wiktor Weibull
Keyword(s):  
Particle Velocity ◽  
Large Scale ◽  
Waveform Inversion ◽  
Transversely Isotropic ◽  
Reconstruction Method ◽  
Full Waveform Inversion ◽  
Reverse Time ◽  
Full Waveform ◽  
Time Migration ◽  
Reconstruction Methods

In reverse time migration (RTM) or full-waveform inversion (FWI), forward and reverse time propagating wavefields are crosscorrelated in time to form either the image condition in RTM or the misfit gradient in FWI. The crosscorrelation condition requires both fields to be available at the same time instants. For large-scale 3D problems, it is not possible, in practice, to store snapshots of the wavefields during forward modeling due to extreme storage requirements. We have developed an approximate wavefield reconstruction method that uses particle velocity field recordings on the boundaries to reconstruct the forward wavefields during the computation of the reverse time wavefields. The method is computationally effective and requires less storage than similar methods. We have compared the reconstruction method to a boundary reconstruction method that uses particle velocity and stress fields at the boundaries and to the optimal checkpointing method. We have tested the methods on a 2D vertical transversely isotropic model and a large-scale 3D elastic FWI problem. Our results revealed that there are small differences in the results for the three methods.

scholarly journals Fall detection for multiple pedestrians using a PCA approach to 3-D inclination

2019 ◽  
Vol 11 ◽  
pp. 184797901987897
Author(s):  
Chien-Feng Wu ◽  
Shir-Kuan Lin
Keyword(s):  
Developing Countries ◽  
Fall Detection ◽  
Experimental Results ◽  
Research Interest ◽  
Depth Camera ◽  
Reconstruction Method ◽  
Object Reconstruction ◽  
Aging Populations ◽  
Early Fall ◽  
Near Future

Aging populations in developing countries have driven increased research interest in issues related to “eldercare,” with a particular focus on early fall detection. This article describes an approach to processing images captured using a depth camera to predict the probable inclination of an imminent fall by a pedestrian. First, the 3-D object reconstruction method is explained and realized. Then, the 3-D object inclination analysis is given. Experimental results show prediction rates achieved accuracy up to 98.11%, making this approach promising for use in applications in hospital and home environments in the near future.

scholarly journals Preclinical dosimetry models and the prediction of clinical doses of novel positron emission tomography radiotracers

2020 ◽  
Vol 10 (1) ◽  
Author(s):  
Adam A. Garrow ◽  
Jack P. M. Andrews ◽  
Zaniah N. Gonzalez ◽  
Carlos A. Corral ◽  
Christophe Portal ◽  
...  
Keyword(s):  
Positron Emission Tomography ◽  
Human Studies ◽  
Emission Tomography ◽  
Whole Body ◽  
Preclinical Model ◽  
Reconstruction Method ◽  
Preclinical Research ◽  
Reconstruction Methods ◽  
Positron Emission ◽  
Pet Radiotracers

Abstract Dosimetry models using preclinical positron emission tomography (PET) data are commonly employed to predict the clinical radiological safety of novel radiotracers. However, unbiased clinical safety profiling remains difficult during the translational exercise from preclinical research to first-in-human studies for novel PET radiotracers. In this study, we assessed PET dosimetry data of six 18F-labelled radiotracers using preclinical dosimetry models, different reconstruction methods and quantified the biases of these predictions relative to measured clinical doses to ease translation of new PET radiotracers to first-in-human studies. Whole-body PET images were taken from rats over 240 min after intravenous radiotracer bolus injection. Four existing and two novel PET radiotracers were investigated: [18F]FDG, [18F]AlF-NOTA-RGDfK, [18F]AlF-NOTA-octreotide ([18F]AlF-NOTA-OC), [18F]AlF-NOTA-NOC, [18F]ENC2015 and [18F]ENC2018. Filtered-back projection (FBP) and iterative methods were used for reconstruction of PET data. Predicted and true clinical absorbed doses for [18F]FDG and [18F]AlF-NOTA-OC were then used to quantify bias of preclinical model predictions versus clinical measurements. Our results show that most dosimetry models were biased in their predicted clinical dosimetry compared to empirical values. Therefore, normalization of rat:human organ sizes and correction for reconstruction method biases are required to achieve higher precision of dosimetry estimates.

scholarly journals Reconstruction Techniques in IceCube using Convolutional and Generative Neural Networks

2019 ◽  
Vol 207 ◽  
pp. 05005 ◽  
Author(s):  
Mirco Huennefeld
Keyword(s):  
Neural Networks ◽  
High Energy Physics ◽  
High Energy ◽  
Reconstruction Method ◽  
Network Architectures ◽  
Reconstruction Methods ◽  
Physics Potential ◽  
Reconstruction Performance ◽  
Maximum Likelihood Methods ◽  
First Results

Reliable and accurate reconstruction methods are vital to the success of high-energy physics experiments such as IceCube. Machine learning based techniques, in particular deep neural networks, can provide a viable alternative to maximum-likelihood methods. However, most common neural network architectures were developed for other domains such as image recogntion. While these methods can enhance the reconstruction performance in IceCube, there is much potential for tailored techniques. In the typical physics use-case, many symmetries, invariances and prior knowledge exist in the data, which are not fully exploited by current network architectures. Novel and specialized deep learning based reconstruction techniques are desired which can leverage the physics potential of experiments like IceCube. A reconstruction method using convolutional neural networks is presented which can significantly increase the reconstruction accuracy while greatly reducing the runtime in comparison to standard reconstruction methods in Ice- Cube. In addition, first results are discussed for future developments based on generative neural networks.

Sign in / Sign up

Export Citation Format

Share Document