Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Adding updated background rejection script with support for pred.root #18

Open
wants to merge 3 commits into
base: efficient
Choose a base branch
from
Open

Adding updated background rejection script with support for pred.root #18

Changes from all commits
Commits
Show all changes
3 commits
Select commit Hold shift + click to select a range
1896334
Add files via upload
aaronw5 Sep 8, 2024
3cffa54
Updating background rejection notebook for .root files
aaronw5 Sep 8, 2024
b4069d7
Adding background rejection with output root files
aaronw5 Sep 8, 2024
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
145 changes: 87 additions & 58 deletions notebooks/backgroundRejection.ipynb
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -1,100 +1,129 @@
{
"cells": [
{
"cell_type": "markdown",
"id": "e2401b40-915a-4fd6-8c4f-d81bd9973f93",
"metadata": {},
"source": [
"## Convert root files into numpy array"
]
},
{
"cell_type": "code",
"execution_count": 1,
"id": "fab67872-f3b8-44ff-b327-1ea3a53c6c71",
"execution_count": 27,
"id": "4a184fac-8f9c-4c63-bff3-0d230eecd71d",
"metadata": {},
"outputs": [],
"source": [
"import os\n",
"import uproot\n",
"import numpy as np\n",
"from sklearn.metrics import accuracy_score\n",
"import matplotlib.pyplot as plt\n",
"import numpy as np\n",
"from sklearn.metrics import roc_curve, auc, roc_auc_score\n",
"from scipy.special import softmax\n",
"\n"
"\n",
"\n",
"# Define the directory path\n",
"directory = \"LinformerEval\"\n",
"labelList = []\n",
"predList = []\n",
"# Loop through the files in the directory\n",
"for filename in os.listdir(directory):\n",
" if filename.endswith(\".root\"):\n",
" try:\n",
" filepath = os.path.join(directory, filename)\n",
" background = uproot.open(filepath)\n",
" tree = background['Events;1']\n",
" df =tree.arrays(library=\"pd\")\n",
" labels = df[['label_QCD', 'label_Hbb', 'label_Hcc', 'label_Hgg', 'label_H4q',\n",
" 'label_Hqql', 'label_Zqq', 'label_Wqq', 'label_Tbqq', 'label_Tbl']]\n",
" pred = df[['score_label_QCD', 'score_label_Hbb', 'score_label_Hcc', 'score_label_Hgg', 'score_label_H4q',\n",
" 'score_label_Hqql', 'score_label_Zqq', 'score_label_Wqq', 'score_label_Tbqq', 'score_label_Tbl']] \n",
" labelList.append(labels.to_numpy())\n",
" predList.append(pred.to_numpy())\n",
" except:\n",
" continue\n",
" \n",
"labels = np.concatenate(labelList).astype(int)\n",
"y_prob = np.concatenate(predList)"
]
},
{
"cell_type": "markdown",
"id": "18a1583c-9c4c-4c59-a824-b89268f4665c",
"metadata": {},
"source": [
"## Find background rejection\n"
]
},
{
"cell_type": "code",
"execution_count": 23,
"id": "47b70f51-9661-4b04-a9fa-cc80ab49d046",
"execution_count": null,
"id": "b8d7f111-e095-4e5a-89f5-e0ada0647b3a",
"metadata": {},
"outputs": [
{
"name": "stdout",
"output_type": "stream",
"text": [
"Rejection at 50.0% for label_Hbb: 4246.284501061571\n",
"Rejection at 50.0% for label_Hcc: 1046.0251046025105\n",
"Rejection at 50.0% for label_Hgg: 82.05128205128204\n",
"Rejection at 50.0% for label_H4q: 416.49312786339027\n",
"Rejection at 99.0% for label_Hqql: 570.7762557077625\n",
"Rejection at 50.0% for label_Zqq: 189.48365703458077\n",
"Rejection at 50.0% for label_Wqq: 210.1281781886951\n",
"Rejection at 50.0% for label_Tbqq: 2347.417840375587\n",
"Rejection at 99.5% for label_Tbl: 1307.1895424836603\n",
"Overall ROC AUC = 0.9357, Accuracy = 0.7969\n"
]
}
],
"outputs": [],
"source": [
"label_list = ['label_QCD' , 'label_Hbb', 'label_Hcc', 'label_Hgg', 'label_H4q', 'label_Hqql', 'label_Zqq', 'label_Wqq', 'label_Tbqq', 'label_Tbl'] # Replace with your actual class labels\n",
"\n",
"n_classes = 10 \n",
"\n",
"predictions = np.load('outputs_base.npy')\n",
"\n",
"labels = np.load('labels_base.npy')\n",
"\n",
"y_prob = softmax(predictions, axis=1) \n",
"\n",
"scores = y_prob[:,1:10]/ (y_prob[:,0][:, np.newaxis] + y_prob[:,1:10])\n",
"\n",
"scores = np.concatenate((y_prob[:,0].reshape(len(scores),1), scores), axis = 1)\n",
"\n",
"rejections = []\n",
"\n",
"for i in range(1, n_classes): \n",
" percent = 0.5\n",
" \n",
" mask = (labels[:, 0] == 1) | (labels[:, i] == 1)\n",
" filtered_labels = labels[mask]\n",
" filtered_scores = scores[mask]\n",
" \n",
" binary_labels = (filtered_labels[:, i] == 1).astype(int)\n",
" \n",
" binary_scores = filtered_scores[:, i]\n",
" try:\n",
" percent = 0.5\n",
" \n",
" mask = (labels[:, 0] == 1) | (labels[:, i] == 1)\n",
" filtered_labels = labels[mask]\n",
" filtered_scores = scores[mask]\n",
" \n",
" binary_labels = (filtered_labels[:, i] == 1).astype(int)\n",
" \n",
" binary_scores = filtered_scores[:, i]\n",
" \n",
" fpr, tpr, thresholds = roc_curve(binary_labels, binary_scores)\n",
" \n",
" fpr, tpr, thresholds = roc_curve(binary_labels, binary_scores)\n",
"\n",
" if i == 5:\n",
" percent = 0.99\n",
" if i == 9:\n",
" percent = 0.995\n",
" \n",
" idx = np.abs(tpr - percent).argmin()\n",
" \n",
" if fpr[idx] != 0:\n",
" rejection = 1 / fpr[idx]\n",
" else:\n",
" rejection = np.inf \n",
" \n",
" rejections.append(rejection)\n",
"\n",
" \n",
" print(f'Rejection at {percent*100}% for {label_list[i]}: {rejection}')\n",
" if i == 5:\n",
" percent = 0.99\n",
" if i == 9:\n",
" percent = 0.995\n",
" \n",
" idx = np.abs(tpr - percent).argmin()\n",
" \n",
" if fpr[idx] != 0:\n",
" rejection = 1 / fpr[idx]\n",
" else:\n",
" rejection = np.inf \n",
" \n",
" rejections.append(rejection)\n",
" \n",
" \n",
" print(f'Rejection at {percent*100}% for {label_list[i]}: {rejection}')\n",
" except:\n",
" continue\n",
"print(labels.shape)\n",
"overall_roc_auc = roc_auc_score(labels, scores, average='macro', multi_class='ovo')\n",
"\n",
"predicted_labels = np.argmax(softmax(base, axis=1), axis=1) \n",
"predicted_labels = np.argmax(y_prob, axis=1) \n",
"true_labels = np.argmax(labels, axis=1) \n",
"\n",
"accuracy = accuracy_score(true_labels, predicted_labels)\n",
"\n",
"print(f'Overall ROC AUC = {overall_roc_auc:.4f}, Accuracy = {accuracy:.4f}')\n"
]
},
{
"cell_type": "code",
"execution_count": null,
"id": "797b4e11-7951-4762-a4e2-592ce9f8db4f",
"metadata": {},
"outputs": [],
"source": []
}
],
"metadata": {
Expand Down