Corpus ID: 67856716

Bayesian Anomaly Detection and Classification

  title={Bayesian Anomaly Detection and Classification},
  author={Ethan Roberts and Bruce A. Bassett and Michelle Lochner},
Statistical uncertainties are rarely incorporated in machine learning algorithms, especially for anomaly detection. Here we present the Bayesian Anomaly Detection And Classification (BADAC) formalism, which provides a unified statistical approach to classification and anomaly detection within a hierarchical Bayesian framework. BADAC deals with uncertainties by marginalising over the unknown, true, value of the data. Using simulated data with Gaussian noise, BADAC is shown to be superior to… Expand
A flexible framework for anomaly Detection via dimensionality reduction
The flexibility of the DRAMA framework allows for significant optimization once some examples of anomalies are available, making it ideal for online anomaly detection, active learning, and highly unbalanced datasets. Expand
Supplier Impersonation Fraud Detection using Bayesian Inference
This paper introduces ProbaSIF, a supplier impersonation fraud detection system that relies on a Bayesian model to perform the classification of a new transaction as legitimate or fraudulent, and demonstrates that the classification system was accurate for a wide set of different clients. Expand
A Knowledge-Based Cognitive Architecture Supported by Machine Learning Algorithms for Interpretable Monitoring of Large-Scale Satellite Networks
This research proposes a knowledge-based cognitive architecture supported by machine learning algorithms for monitoring satellite network traffic capable of supporting and augmenting infrastructure engineers in finding and understanding the causes of faults in network through the fusion of the results of machine learning models and rules derived from human domain experience. Expand
Data Challenges as a Tool for Time-domain Astronomy
  • R. Hložek
  • Physics, Computer Science
  • Publications of the Astronomical Society of the Pacific
  • 2019
This work highlights some recent time-domain challenges, with particular focus on the Photometric LSST Astronomical Time series Classification Challenge (PLAsTiCC), and describes metrics used to evaluate the performance of those entering data challenges. Expand
Hack The Box: Fooling Deep Learning Abstraction-Based Monitors
It is demonstrated that novelty detection itself ends up as an attack surface when crafting adversarial samples that fool the deep learning classifier and bypass the novelty detection monitoring at the same time. Expand
Conceptual Model of an Intelligent System for Detecting Fraud During Mobile Applications Installation
  • T. Polhul
  • Computer Science
  • 2019 10th International Conference on Dependable Systems, Services and Technologies (DESSERT)
  • 2019
The conceptual model of an intelligent system for fraud detection during mobile applications installation allows processing heterogeneous input data, which gives the opportunity to form a generalized fraudsters fingerprint in the process. Expand
Astronomaly: Personalised active anomaly detection in astronomical data
Astronomaly is introduced: a general anomaly detection framework with a novel active learning approach designed to provide personalised recommendations that can operate on most types of astronomical data, including images, light curves and spectra. Expand
A Flexible Framework for Anomaly Detection via Dimensionality Reduction
  • A. V. Sadr, B. Bassett, M. Kunz
  • Computer Science, Physics
  • 2019 6th International Conference on Soft Computing & Machine Intelligence (ISCMI)
  • 2019
The flexibility of the DRAMA framework allows for significant optimization once some examples of anomalies are available, making it ideal for online anomaly detection, active learning and highly unbalanced datasets. Expand


Isolation Forest
  • F. Liu, K. Ting, Z. Zhou
  • Computer Science
  • 2008 Eighth IEEE International Conference on Data Mining
  • 2008
The use of isolation enables the proposed method, iForest, to exploit sub-sampling to an extent that is not feasible in existing methods, creating an algorithm which has a linear time complexity with a low constant and a low memory requirement. Expand
Parameter estimation with Bayesian estimation applied to multiple species in the presence of biases and correlations
The original formulation of Bayesian estimation applied to multiple species (BEAMS) showed how to use a data set contaminated by points of multiple underlying types to perform unbiased parameterExpand
LOF: identifying density-based local outliers
This paper contends that for many scenarios, it is more meaningful to assign to each object a degree of being an outlier, called the local outlier factor (LOF), and gives a detailed formal analysis showing that LOF enjoys many desirable properties. Expand
A Bayesian Approach to Classifying Supernovae With Color
Upcoming large-scale ground- and space- based supernova surveys will face a challenge identifying supernova candidates largely without the use of spectroscopy. Over the past several years, a numberExpand
Bayesian Methods for Nonlinear Classification and Regression
  • R. Soyer
  • Computer Science, Mathematics
  • Technometrics
  • 2004
The “hint of quantum mechanics” via commuting operators reminded me that I never really understood my undergraduate course in quantum mechanics, and this is not a text from which to learn Bayesian methods. Expand
Bayesian estimation applied to multiple species
Observed data are often contaminated by undiscovered interlopers, leading to biased parameter estimation. Here we present BEAMS (Bayesian estimation applied to multiple species) which significantlyExpand
On the Optimality of the Simple Bayesian Classifier under Zero-One Loss
The Bayesian classifier is shown to be optimal for learning conjunctions and disjunctions, even though they violate the independence assumption, and will often outperform more powerful classifiers for common training set sizes and numbers of attributes, even if its bias is a priori much less appropriate to the domain. Expand
Bayesian Classification With Gaussian Processes
A Bayesian treatment is provided, integrating over uncertainty in y and in the parameters that control the Gaussian process prior the necessary integration over y is carried out using Laplace's approximation, and the method is generalized to multiclass problems (m>2) using the softmax function. Expand
Predicting good probabilities with supervised learning
We examine the relationship between the predictions made by different learning algorithms and true posterior probabilities. We show that maximum margin methods such as boosted trees and boostedExpand
Supernova (SN) cosmology without spectroscopic confirmation is an exciting new frontier, which we address here with the Bayesian Estimation Applied to Multiple Species (BEAMS) algorithm and the fullExpand