stat441F21: Difference between revisions

From statwiki
Jump to navigation Jump to search
(Undo revision 49935 by Aghodsib (talk))
(4 intermediate revisions by 3 users not shown)
Line 4: Line 4:


<!--[https://goo.gl/forms/apurag4dr9kSR76X2 Your feedback on presentations]-->
<!--[https://goo.gl/forms/apurag4dr9kSR76X2 Your feedback on presentations]-->
Project # 0 Group members:
Bowen,Zhang
Laura, Chen
Title: Improving ECG Classification Using Generative Adversarial Networks
Description: Electrocardiography (ECG) is a non-invasive tool used for diagnosis and followup of cardiac anomalies, functional dis-
orders and cardiac arrhythmias. In this paper, the author developed classification and diagnosis of ECG data using long short-term memory algorithm. In addition, the author also used Generative Adversarial Network to generate Electrocardiography (ECG) synthetic data. The studied data contains 48 half-hour ECG record obtained from patients studied by the BIH Arrhythmia Laboratory between 1975 and 1979. Each record contains two 30-minutes ECG lead signals digitized at 360 samples per second. Annotations for both heartbeat class and timing information are included in the dataset.
The LSTM-based Heartbeat Classification considered the temporal nature of ECG signals while avoiding strong model assumptions required by preprocessing and feature engineering approach. When generating ECG heartbeat signals using GANs, the author included generator network and discriminator network to train the model.
The results of the LSTM network classification result for all 4 heartbeat classes outperforms the fully connected network and achieved accuracy all above 80%. For the generated ECG signal, all type of heartbeats the GAN learns to generate synthetic heartbeats which have which have the same PQRST morphology as their corresponding real heartbeats from the training set.


=Paper presentation=
=Paper presentation=
Line 37: Line 22:
|Week of Nov 16 || Ali Ghodsi || ||  ||  || ||
|Week of Nov 16 || Ali Ghodsi || ||  ||  || ||
|-
|-
|Week of Nov 16 || || || ||   || ||
|Week of Nov 16 || Group 10 || || A Probabilistic Approach to Neural Network Pruning || [http://proceedings.mlr.press/v139/qian21a/qian21a.pdf] || ||
|-
|-
|Week of Nov 16 || || || ||   || ||
|Week of Nov 16 || Jared Feng, Xipeng Huang, Mingwei Xu, Tingzhou Yu|| || Don't Just Blame Over-parametrization for Over-confidence: Theoretical Analysis of Calibration in Binary Classification || [http://proceedings.mlr.press/v139/bai21c/bai21c.pdf Paper]  || ||
|-
|-
|Week of Nov 16 || || ||  ||  || ||
|Week of Nov 16 || || ||  ||  || ||
|-
|-

Revision as of 22:49, 15 October 2021


Project Proposal

Paper presentation

Date Name Paper number Title Link to the paper Link to the summary Link to the video
Sep 15 (example) Ri Wang Sequence to sequence learning with neural networks. Paper Summary [1]
Week of Nov 16 Ali Ghodsi
Week of Nov 16 Group 10 A Probabilistic Approach to Neural Network Pruning [2]
Week of Nov 16 Jared Feng, Xipeng Huang, Mingwei Xu, Tingzhou Yu Don't Just Blame Over-parametrization for Over-confidence: Theoretical Analysis of Calibration in Binary Classification Paper
Week of Nov 16