Abstract: In many outlier detection tasks, only training data
belonging to one class, i.e., the positive class, is available. The
task is then to predict a new data point as belonging either to
the positive class or to the negative class, in which case the
data point is considered an outlier. For this task, we propose a
novel corrupted Generative Adversarial Network (CorGAN). In the
adversarial process of training CorGAN, the Generator generates
outlier samples for the negative class, and the Discriminator is trained
to distinguish the positive training data from the generated negative
data. The proposed framework is evaluated using an image dataset
and a real-world network intrusion dataset. Our outlier-detection
method achieves state-of-the-art performance on both tasks.
Abstract: In this paper, an improved method for estimating fundamental matrix is proposed. The method is applied effectively to monocular camera based moving object detection. The method consists of corner points detection, moving object’s motion estimation and fundamental matrix calculation. The corner points are obtained by using Harris corner detector, motions of moving objects is calculated from pyramidal Lucas-Kanade optical flow algorithm. Through epipolar geometry analysis using RANSAC, the fundamental matrix is calculated. In this method, we have improved the performances of moving object detection by using two threshold values that determine inlier or outlier. Through the simulations, we compare the performances with varying the two threshold values.
Abstract: In this paper numerous robust fitting procedures are considered in estimating spatial variograms. In spatial statistics, the conventional variogram fitting procedure (non-linear weighted least squares) suffers from the same outlier problem that has plagued this method from its inception. Even a 3-parameter model, like the variogram, can be adversely affected by a single outlier. This paper uses the Hogg-Type adaptive procedures to select an optimal score function for a rank-based estimator for these non-linear models. Numeric examples and simulation studies will demonstrate the robustness, utility, efficiency, and validity of these estimates.
Abstract: Sentiment analysis means to classify a given review
document into positive or negative polar document. Sentiment
analysis research has been increased tremendously in recent times
due to its large number of applications in the industry and academia.
Sentiment analysis models can be used to determine the opinion of
the user towards any entity or product. E-commerce companies can
use sentiment analysis model to improve their products on the basis
of users’ opinion. In this paper, we propose a new One-class Support
Vector Machine (One-class SVM) based sentiment analysis model
for movie review documents. In the proposed approach, we initially
extract features from one class of documents, and further test the
given documents with the one-class SVM model if a given new test
document lies in the model or it is an outlier. Experimental results
show the effectiveness of the proposed sentiment analysis model.
Abstract: At-site flood frequency analysis is used to estimate
flood quantiles when at-site record length is reasonably long. In
Australia, FLIKE software has been introduced for at-site flood
frequency analysis. The advantage of FLIKE is that, for a given
application, the user can compare a number of most commonly
adopted probability distributions and parameter estimation methods
relatively quickly using a windows interface. The new version of
FLIKE has been incorporated with the multiple Grubbs and Beck test
which can identify multiple numbers of potentially influential low
flows. This paper presents a case study considering six catchments in
eastern Australia which compares two outlier identification tests
(original Grubbs and Beck test and multiple Grubbs and Beck test)
and two commonly applied probability distributions (Generalized
Extreme Value (GEV) and Log Pearson type 3 (LP3)) using FLIKE
software. It has been found that the multiple Grubbs and Beck test
when used with LP3 distribution provides more accurate flood
quantile estimates than when LP3 distribution is used with the
original Grubbs and Beck test. Between these two methods, the
differences in flood quantile estimates have been found to be up to
61% for the six study catchments. It has also been found that GEV
distribution (with L moments) and LP3 distribution with the multiple
Grubbs and Beck test provide quite similar results in most of the
cases; however, a difference up to 38% has been noted for flood
quantiles for annual exceedance probability (AEP) of 1 in 100 for one
catchment. This finding needs to be confirmed with a greater number
of stations across other Australian states.