Novelty detection

Tiitinen, H., P. May, K. Reinikainen, and R. Naatanen. 1994. "Acute Novelty Detection in Humans Is Governed by Preattentive Sensory Memory." Nature 372 90-92. 

Figure 15.1 Figures to describe the basis of novelty and margin detection to provide a confidence measure for K-PLS predictions, (a) Novelty detection, (b) margin detection for classification, and (c) margin detection for regression.

Hristozov, D., Oprea, T.I., Gasteiger, J. Ligand-based virtual screening by novelty detection with self-organizing maps. J. Chem. Inf. Model. 2007,47,2044-62. 

Classification, including pattern and sequence recognition, novelty detection, and sequential decision making... 

Onboard fault detection is such an important facet of an intelligent sensor that density-based novelty detection may be used in parallel with more traditional approaches such as a residual-based fault detection approach [12]. Here, time series predictions from a data-based model using recent measurements retained in a buffer are compared with the actual current measurement provided by the sensor, to calculate a residual error between the two estimates. Significant discrepancy highlighted by a large residual error is indicative of an error condition. 

One-class classification (novelty detection) One-Class Support Vector Machine (1-SVM) [12], Support Vector Data Description (SVDD) [76] Virtual screening based of similarity of molecular fields... 

Markou M, Singh S (2003a) Novelty detection a review—part 1 statistical approaches. Signal Process 83(12) 2481-2497... 

A Hybrid Trial to Trial Wavelet Coherence and Novelty Detection Scheme for a Fast and Clear Notification of Habituation An Objective Uncomfortable Loudness... 

Objective To determine an uncomfortable loudness (UCL) level is not an easy task, especially in children. A need to objectively measure this level is crucial as the age of hearing devices candidates is getting younger. Previous studies have shown that the feasibility of habituation correlates in late auditory evoked potentials for a measurement technique of UCL identification is promising. Nevertheless, a scheme that could provide a fast and clear notification of an UCL level is reached is desirable. The present study has introduced a hybrid trial to trial wavelet coherence and novelty detection scheme to extract and to notify objectively the habituation correlates in late auditory evoked potentials. 

Keywords— habituation, wavelet coherence, novelty detection, late auditory evoked potential... 

In this paper, the post-processing in [5] is inqrroved by combining the SSTC scheme with the novelty detection approach in order to achieve a fast and a reliable identification scheme for habituation extractiom Similar data from [5] is used and the comparison between the previous and present of post processing results in made. The details the present technique and a brief review of data acquisition are discussed in the following sections. 

Trial to trial provides informative features that describe habituate response or non-habituate response. In order to extract the moment of the habituation presence, we further analyze the wavelet coherence results with novelty detection. We have observed in [5] that reduction of wavelet coherence (habituation) occurs after several trials after the experiment began. Therefore, we take these first q trials (training set) to train a classifier. This classifier is generated by all information of q trials and forms a hypothetical sphere with center c and radius K. To minimize this sphere so that the learning machine free from false negative errors, the problem is solved by... 

In the present study, we further analyze the results in [5] in order to achieve a fast and rehable identification of habituate response. With a combination of trial to trial wavelet coherence and novelty detection schemes, identification of habituation moment is clearly highlighted. Trial to trial wavelet coherence is a reliable method to extract habituation and novelty detection analysis is able to highlight the abnormality in a train of data which provide a suitable method for identifying a reduction of response in real time. 

In the present study, we improved post processing technique to extract habituation in LAEPs by combining trial to trial wavelet coherence with novelty detection. The identification of habituation presence is faster and clearly hig-hhghted in comparison to previous study. These findings, give a promising technique for objectively determine UCL procedure and analysis. 

---