A deep learning model (CNN) build to classify Quark and Gluon using particle images.
The dataset was provided by the CERN (2 https://cernbox.cern.ch/index.php/s/hqz8zE7oxyPjvsL). The file name used here is: 'QCDToGGQQ_IMGjet_RH1all_jet0_run0_n36272.test.snappy.parquet' where the 'X_jet' was the input image and 'y' was the target.
-
For training set:
- Input: 10,000 images of size 125x125 and 3 channels
- Target: 10,000 values of 0 or 1
-
For validation set:
- Input: 5390 images of size 125x125 and 3 channels
- Target: 5390 values of 0 or 1
The model and code description is mentioned in the table below:
| DL Framework | Keras |
| Keras version | 2.4.3 |
| Libraries | Numpy, pandas, pyarrow |
| Pre-trained weights | Imagenet |
| Activation functions | ReLU, sigmoid |
| Optimizer | RMSprop |
| Learning Rate | 0.0001 |
| Batch size | 32 |
| Epochs | 100 |
| Loss function | Binary Crossentropy |
| Regularizer | Dropout |
| Metrics | Accuracy |