A deep learning model (CNN) build to estimate/regress the mass of the particle based on particle images.
The dataset was provided by the CERN. The file name used here is: 'E2E_Regression.parquet.9' where the 'X_jet' was the input image and 'am' was the mass of the particle used as target.
-
For training set:
- Input: 5139 images of size 125x125 and 4 channels
- Target: 5139 values
-
For validation set:
- Input: 1284 images of size 125x125 and 4 channels
- Target: 1284 values
The model and code description is mentioned in the table below:
| DL Framework | Keras |
| Keras version | 2.4.3 |
| Libraries | Numpy, pandas, pyarrow |
| Layers | Conv2D, Batchnormalization, Flatten, Dense |
| Activation functions | ReLU, Linear |
| Optimizer | Adam |
| Learning Rate | 0.0001 |
| Batch size | 32 |
| Epochs | 90 |
| Loss function | Mean Squared error (mse) |
| Metrics | Mean Absolute Percentage Error (mape), Mean Squared Error (mse), Mean Absolute Error (mae) |
| Dataset | Loss | mape | mse | mae |
| Train set | 0.0030 | 1.1389 | 0.0030 | 0.0436 |
| Validation set | 0.0483 | 4.8653 | 0.0483 | 0.1885 |