Changes for page Xabier Cid Vidal

Summary

• Page properties (2 modified, 0 added, 0 removed)

Details

Page properties

Title

...	...	@@ -1,1 +1,1 @@
1		-CidVidal_Xabier
	1	+Xabier Cid Vidal

Content

...	...	@@ -1,23 +1,26 @@
1		-=== Creating classifiers to detect LLPs at LHCb
2		-===
	1	+=== Creating classifiers to detect LLPs at LHCb ===
3	3
4		-Several Beyond the Standard Model theories suggest that Long-Lived Particles (LLPs) might mostly evade current particle physics detectors but have a small chance of decaying into known elementary particles, such as muons, making them potentially detectable.
	3	+=== ===
	4	+
	5	+Several Beyond the Standard Model theories suggest that Long-Lived Particles (LLPs) might mostly evade current particle physics detectors but have a small chance of decaying into known elementary particles, such as muons, making them potentially detectable.
5	5	The LHCb detector offers great precision in detecting production/decay vertices. Therefore, using Machine Learning tools supported by these measurements is considered an enhancing element, improving the sensitivity to these models. For this project we will develop such tools, comparing the performance of different algorithms, such as AdaBoost and GradientBoost.
6	6
7	7
8	8
9		-=== Finding the best classifier to observe eta' -> mu mu at LHCb
10		-===
	10	+=== Finding the best classifier to observe eta' -> mu mu at LHCb ===
11	11
	12	+=== ===
	13	+
12	12	Decays of the pseudoscalar resonances eta and eta' could turn out to be an important factor in fine-tuning the theoretical predictions for g_mu -2, one of the larger discrepancies between the Standard Model of Particle Physics and experimental measurements. In particular, eta' -> mu mu has not been observed yet, and doing so could represent a landmark achievement for the field, as well as the experiment that does it.
13		-An ongoing analysis of data collected by the LHCb experiment at CERN is showing promising signs of observing evidence of it with a significance higher than 3 sigmas. It is also shown that a key factor to improve this sensitivity could be a better-performing multivariate classifier, that allows us to distinguish between background-like and signal-like data. In this project, we will be fine-tuning the classifier to achieve an optimal performance.
	15	+An ongoing analysis of data collected by the LHCb experiment at CERN is showing promising signs of observing evidence of it with a significance higher than 3 sigmas. It is also shown that a key factor to improve this sensitivity could be a better-performing multivariate classifier, that allows us to distinguish between background-like and signal-like data. In this project, we will be fine-tuning the classifier to achieve an optimal performance.
14	14	To do so, we will be testing multiple classifying algorithms from Python, including XGBoost, AdaBoost, and even neural networks that could take us over the top and achieve an area under the ROC curve of 0.99, improving upon the current result of 0.98.
15	15
16	16
17	17
18		-=== Optimizing hyperparameters in classifiers to observe eta' -> mu mu at LHCb
19		-===
	20	+=== Optimizing hyperparameters in classifiers to observe eta' -> mu mu at LHCb ===
20	20
	22	+=== ===
	23	+
21	21	The decays of the pseudoscalar resonances eta and eta' may play a crucial role in refining theoretical predictions for g_mu - 2, which is one of the significant discrepancies between the Standard Model of Particle Physics and experimental measurements. Notably, the decay eta' -> mu mu has not yet been observed. Achieving this observation would be a landmark accomplishment for the field and the experiment that accomplishes it.
22	22	Current analysis of data collected by the LHCb experiment at CERN shows promising signs of observing evidence of this decay with a significance greater than 3 sigmas. Improving this sensitivity could significantly benefit from a more effective multivariate classifier, which can better differentiate between background-like and signal-like data. This project aims to fine-tune the classifier to achieve optimal performance.
23	23	For this, we will be trying with different input variables in the dataset, testing how much each one contributes to the overall result. We will also fine-tune the hyperparameters of the training algorithm using the Python optimization framework Optuna, which allows us to efficiently test different strategies and choose the best result.