Classification without labels: learning from mixed samples in high energy physics

Metodiev, Eric M.; Nachman, B. P.; Thaler, Jesse

doi:10.1007/jhep10(2017)174

Cited by 205 publications

(203 citation statements)

References 88 publications

Supporting

Mentioning

202

Contrasting

Order By: Relevance

“…Part of these approaches use machine learning techniques, which is another direction into which new physics searches at the LHC can expand, as has been also proposed recently in several other contexts (e.g., refs. [34][35][36][37][38][39][40][41][42][43]).…”

Section: Discussionmentioning

confidence: 99%

Searching for periodic signals in kinematic distributions using continuous wavelet transforms

Beauchesne

Kats

2020

Eur. Phys. J. C

View full text Add to dashboard Cite

Many models of physics beyond the Standard Model include towers of particles whose masses follow an approximately periodic pattern with little spacing between them. These resonances might be too weak to detect individually, but could be discovered as a group by looking for periodic signals in kinematic distributions. The continuous wavelet transform, which indicates how much a given frequency is present in a signal at a given time, is an ideal tool for this. In this paper, we present a series of methods through which continuous wavelet transforms can be used to discover periodic signals in kinematic distributions. Some of these methods are based on a simple test statistic, while others make use of machine learning techniques. Some of the methods are meant to be used with a particular model in mind, while others are model-independent. We find that continuous wavelet transforms can give bounds comparable to current searches and, in some cases, be sensitive to signals that would go undetected by standard experimental strategies.

show abstract

Section: Discussionmentioning

confidence: 99%

Searching for periodic signals in kinematic distributions using continuous wavelet transforms

Beauchesne

Kats

2020

Eur. Phys. J. C

View full text Add to dashboard Cite

show abstract

“…For example, this was used in the γγ channel of the recent tth observation, [19,20] and General Search [21][22][23] strategies are from CMS and ATLAS, respectively. LDA stands for Latent Dirichlet Allocation [37,78], ANOmaly detection with Density Estimation (ANODE) is the method presented in this paper, and CWoLa stands for Classification Without Labels [32,33,77]. Direct density estimation is a form of side-banding where the multidimensional feature space density is learned conditional on the resonant feature (see Sec.…”

Section: Bsm Sensitivitymentioning

confidence: 99%

“…Instead, CWoLa hunting uses neural networks to identify differences between signal regions and neighboring sideband regions. By turning the problem into a supervised learning task [77], CWoLa is able to effectively find rare resonant signals. However, CWoLa hunting has certain requirements on the independence of the discriminating features and the resonant feature.…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Anomaly detection with density estimation

2020

Self Cite

View full text Add to dashboard Cite

We leverage recent breakthroughs in neural density estimation to propose a new unsupervised anomaly detection technique (ANODE). By estimating the probability density of the data in a signal region and in sidebands, and interpolating the latter into the signal region, a likelihood ratio of data vs. background can be constructed. This likelihood ratio is broadly sensitive to overdensities in the data that could be due to localized anomalies. In addition, a unique potential benefit of the ANODE method is that the background can be directly estimated using the learned densities. Finally, ANODE is robust against systematic differences between signal region and sidebands, giving it broader applicability than other methods. We demonstrate the power of this new approach using the LHC Olympics 2020 R&D Dataset. We show how ANODE can enhance the significance of a dijet bump hunt by up to a factor of 7 with a 10% accuracy on the background prediction. While the LHC is used as the recurring example, the methods developed here have a much broader applicability to anomaly detection in physics and beyond.

show abstract

“…6 A large effort is made to address this aspect of phenomenology thanks to new data analysis techniques. For example, new interesting results are obtained thanks to the machine learning approach [39,40]. Figure 5 shows the spin weight histograms for the H and X samples.…”

Section: Spin Dependent Characteristicsmentioning

confidence: 99%

Production of $$\tau \tau jj$$ττjj final states at the LHC and the TauSpinner algorithm: the spin-2 case

et al. 2018

View full text Add to dashboard Cite

The TauSpinner algorithm is a tool that allows one to modify the physics model of the Monte Carlo generated samples due to the changed assumptions of event production dynamics, but without the need of re-generating events. With the help of weights τ -lepton production or decay processes can be modified accordingly to a new physics model. In a recent paper a new version TauSpinner ver.2.0.0 has been presented which includes a provision for introducing non-standard states and couplings and study their effects in the vector-boson-fusion processes by exploiting the spin correlations of τ -lepton pair decay products in processes where final states include also two hard jets. In the present paper we document how this can be achieved taking as an example the non-standard spin-2 state that couples to Standard Model particles and tree-level matrix elements with complete helicity information included for the parton-parton scattering amplitudes into a τ -lepton pair and two outgoing partons. This implementation is prepared as the external (user-provided) routine for the TauSpinner algorithm. It exploits amplitudes generated by MadGraph5 and adapted to the TauSpinner algorithm format. Consistency tests of the implemented matrix elements, re-weighting algorithm and numerical results for observables sensitive to τ polarisation are presented.

show abstract

Classification without labels: learning from mixed samples in high energy physics

Cited by 205 publications

References 88 publications

Searching for periodic signals in kinematic distributions using continuous wavelet transforms

Searching for periodic signals in kinematic distributions using continuous wavelet transforms

Anomaly detection with density estimation

Production of $$\tau \tau jj$$ττjj final states at the LHC and the TauSpinner algorithm: the spin-2 case

Contact Info

Product

Resources

About