Current Status on Shower Clustering Improvements
Description: Current Status on Shower Clustering Improvements Seeing the success of the CNN encoding, we are changing hyperparameters of the CNN to test for potential improvements Specifically, we are changing the number of filters, the depth of the
Related Topics
Download Presentation
"Current Status on Shower Clustering Improvements" is the property of its rightful owner. Permission is granted to download and print the materials on this website for personal, non-commercial use only, and to display it on your personal computer provided you do not modify the materials and that you retain all copyright notices contained in the materials. By downloading content from our website, you accept the terms of this agreement.
Presentation Transcript
slide1. Current Status on Shower Clustering Improvements Seeing the success of the CNN encoding, we are changing hyperparameters of the CNN to test for potential improvements
Specifically, we are changing the number of filters, the depth of the CNN, and the number of repetitions. Additionally testing different pooling methods and combination edge and node cnn encoder.
Running now with 4 GPU per job: have 50+ epochs for nom, rep2, rep3, rep4, pool mode avg, pool mode max; started on filter runs.
We are also running depth 3,5,7 cases (50 epochs by next week). Z. Djurcic et al. 1<br>
slide2. nominal
avg
max 2 Preliminary Inference: pooling modes (avg, max), vs nom ARI Purity Efficiency<br>
slide3. Edge Accuracy and Loss Total Accuracy and Loss 3 Preliminary Inference: pooling modes (avg, max), vs nom<br>
slide4. nominal
rep2
rep3
rep4 4 ARI Efficiency Purity Preliminary Inference: rep2, rep3, rep4, vs nom<br>
slide5. Edge Accuracy and Loss Total Accuracy and Loss 5 Preliminary Inference: rep2, rep3, rep4, vs nom nominal
rep2
rep3
rep4<br>
Specifically, we are changing the number of filters, the depth of the CNN, and the number of repetitions. Additionally testing different pooling methods and combination edge and node cnn encoder.
Running now with 4 GPU per job: have 50+ epochs for nom, rep2, rep3, rep4, pool mode avg, pool mode max; started on filter runs.
We are also running depth 3,5,7 cases (50 epochs by next week). Z. Djurcic et al. 1<br>
slide2. nominal
avg
max 2 Preliminary Inference: pooling modes (avg, max), vs nom ARI Purity Efficiency<br>
slide3. Edge Accuracy and Loss Total Accuracy and Loss 3 Preliminary Inference: pooling modes (avg, max), vs nom<br>
slide4. nominal
rep2
rep3
rep4 4 ARI Efficiency Purity Preliminary Inference: rep2, rep3, rep4, vs nom<br>
slide5. Edge Accuracy and Loss Total Accuracy and Loss 5 Preliminary Inference: rep2, rep3, rep4, vs nom nominal
rep2
rep3
rep4<br>