Semi-supervised detection of collective anomalies with an application in high energy particle physics

Vatanen, Tommi; Kuusela, Mikael; Malmi, Eric; Raiko, Tapani; Aaltonen, T.; Nagai, Y.

doi:10.1109/ijcnn.2012.6252712

Cited by 19 publications

(21 citation statements)

References 18 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…We find only the study by Vatanen [16] as the closest one to our approach. Although [16] used semisupervised collective anomaly detection method considering the distribution differences, the main purpose was to classify the anomalies. The authors did not give any results for finding the parameters of the distribution.…”

Section: Estimation Of Distribution Parametersmentioning

confidence: 98%

“…Since the distribution of the background is known in this problem, semisupervised methods are able to perform better than the unsupervised ones. Semisupervised methods generally assume that only normal data have labeling information [5,16,17]. These methods attempt to estimate the model of the normal data by using the training set and classify each data of the unlabeled test set as an anomaly if it seems unlikely to have been proposed by the process corresponding to the normal data distribution.…”

Section: Introductionmentioning

confidence: 99%

“…These methods attempt to estimate the model of the normal data by using the training set and classify each data of the unlabeled test set as an anomaly if it seems unlikely to have been proposed by the process corresponding to the normal data distribution. In [16], a semisupervised anomaly detection method is proposed and applied to the signal classification problem in high energy physics. In this method, a mixture of Gaussians is fitted to a labeled background data, and then a mixture of this background model and a number of additional Gaussians are fitted to an unlabeled sample containing both background and signal.…”

Section: Introductionmentioning

confidence: 99%

See 2 more Smart Citations

Importance-based signal detection and parameter estimation with applications tonew particle search

Doǧan¹,

Sönmez²,

Demir³

2019

Turk J Elec Eng & Comp Sci

View full text Add to dashboard Cite

One of the hardest challenges in data analysis is perhaps the detection of rare anomalous data buried in a huge normal background. We study this problem by constructing a novel method, which is a combination of the Kullback-Leibler importance estimation procedure based anomaly detection algorithm and linear discriminant classifier. We choose to illustrate it with the example of charged Higgs boson (CHB) search in particle physics. Indeed, the Large Hadron Collider experiments at CERN ensure that CHB signal must be a tiny effect within the irreducible W-boson background. In simulations, different CHB events with different characteristics are produced and judiciously mixed with the non-CHB data, and the proposed method is applied. Our results show that distribution parameters of weak CHB signals can be estimated with high performance. This anomaly detection method is general enough to apply to similar problems in other fields (e.g., astrophysical, medical, engineering problems).

show abstract

Section: Estimation Of Distribution Parametersmentioning

confidence: 98%

Section: Introductionmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Importance-based signal detection and parameter estimation with applications tonew particle search

Doǧan¹,

Sönmez²,

Demir³

2019

Turk J Elec Eng & Comp Sci

View full text Add to dashboard Cite

show abstract

“…Importantly, FNF is agnostic to the concrete density estimation method (as we will show in Section 6), and can thus benefit from future advances in the field. Finally, we note that density estimation has been successfully employed in a variety of security-critical areas such as fairness [7], adversarial robustness [52], and anomaly detection [53][54][55][56][57][58][59][60][61][62][63][64].…”

Section: Algorithm 1 Learning Fair Normalizing Flowsmentioning

confidence: 99%

Fair Normalizing Flows

Balunović¹,

Ruoss²,

Vechev³

2021

Preprint

View full text Add to dashboard Cite

Fair representation learning is an attractive approach that promises fairness of downstream predictors by encoding sensitive data. Unfortunately, recent work has shown that strong adversarial predictors can still exhibit unfairness by recovering sensitive attributes from these representations. In this work, we present Fair Normalizing Flows (FNF), a new approach offering more rigorous fairness guarantees for learned representations. Specifically, we consider a practical setting where we can estimate the probability density for sensitive groups. The key idea is to model the encoder as a normalizing flow trained to minimize the statistical distance between the latent representations of different groups. The main advantage of FNF is that its exact likelihood computation allows us to obtain guarantees on the maximum unfairness of any potentially adversarial downstream predictor. We experimentally demonstrate the effectiveness of FNF in enforcing various group fairness notions, as well as other attractive properties such as interpretability and transfer learning, on a variety of challenging real-world datasets.Preprint. Under review.

show abstract

“…This is a typical feature of new class detection. The recent paper by Vatanen et al (2012), which was motivated by a real data analysis problem of experimental high energy physics, overcomes this difficulty. A two-step procedure is proposed.…”

Section: Introductionmentioning

confidence: 99%

Margin-free classification and new class detection using finite Dirichlet mixtures

John,

Brazzale,

Süveges

2021

Preprint

View full text Add to dashboard Cite

We present a margin-free finite mixture model which allows us to simultaneously classify objects into known classes and to identify possible new object types using a set of continuous attributes. This application is motivated by the needs of identifying and possibly detecting new types of a particular kind of stars known as variable stars. We first suitably transform the physical attributes of the stars onto the simplex to achieve scale invariance while maintaining their dependence structure. This allows us to compare data collected by different sky surveys which can have different scales. The model hence combines a mixture of Dirichlet mixtures to represent the known classes with the semi-supervised classification strategy of Vatanen et al. ( 2012) for outlier detection. In line with previous work on semiparametric model-based clustering, the single Dirichlet distributions can be seen as providing the baseline pattern of the data. These are then combined to effectively model the complex distributions of the attributes for the different classes. The model is estimated using a hierarchical two-step procedure which combines a suitably adapted version of the Expectation-Maximization (EM) algorithm with Bayes' rule. We validate our model on a reliable sample of periodic variable stars available in the literature (Dubath et al., 2011) achieving an overall classification accuracy of 71.95% and a sensitivity of 86.11% and a specificity of 99.79% for new class detection.

show abstract

Semi-supervised detection of collective anomalies with an application in high energy particle physics

Cited by 19 publications

References 18 publications

Importance-based signal detection and parameter estimation with applications tonew particle search

Importance-based signal detection and parameter estimation with applications tonew particle search

Fair Normalizing Flows

Margin-free classification and new class detection using finite Dirichlet mixtures

Contact Info

Product

Resources

About