scholarly journals Mining Hard Negative Samples for SAR-Optical Image Matching Using Generative Adversarial Networks

2018 ◽  
Vol 10 (10) ◽  
pp. 1552 ◽  
Author(s):  
Lloyd Hughes ◽  
Michael Schmitt ◽  
Xiao Zhu
Keyword(s):  
Image Matching ◽  
False Positive Rate ◽  
Optical Image ◽  
Training Data ◽  
Generative Adversarial Networks ◽  
Matching Task ◽  
Matching Network ◽  
Image Patches ◽  
Positive Rate ◽  
Real Imagery

In this paper, we propose a generative framework to produce similar yet novel samples for a specified image. We then propose the use of these images as hard-negatives samples, within the framework of hard-negative mining, in order to improve the performance of classification networks in applications which suffer from sparse labelled training data. Our approach makes use of a variational autoencoder (VAE) which is trained in an adversarial manner in order to learn a latent distribution of the training data, as well as to be able to generate realistic, high quality image patches. We evaluate our proposed generative approach to hard-negative mining on a synthetic aperture radar (SAR) and optical image matching task. Using an existing SAR-optical matching network as the basis for our investigation, we compare the performance of the matching network trained using our approach to the baseline method, as well as to two other hard-negative mining methods. Our proposed generative architecture is able to generate realistic, very high resolution (VHR) SAR image patches which are almost indistinguishable from real imagery. Furthermore, using the patches as hard-negative samples, we are able to improve the overall accuracy, and significantly decrease the false positive rate of the SAR-optical matching task—thus validating our generative hard-negative mining approaches’ applicability to improve training in data sparse applications.

Supervised Learning in Absence of Accurate Class Labels

2017 ◽  
pp. 220-233
Author(s):  
Ramasubramanian Sundararajan ◽  
Hima Patel ◽  
Manisha Srivastava
Keyword(s):  
Supervised Learning ◽  
False Positive Rate ◽  
Ensemble Methods ◽  
Optimization Method ◽  
Training Data ◽  
Optimal Separation ◽  
Supervised Learning Algorithms ◽  
Positive Class ◽  
Positive Rate ◽  
Class Labels

Traditionally supervised learning algorithms are built using labeled training data. Accurate labels are essential to guide the classifier towards an optimal separation between the classes. However, there are several real world scenarios where the class labels at an instance level may be unavailable or imprecise or difficult to obtain, or in situations where the problem is naturally posed as one of classifying instance groups. To tackle these challenges, we draw your attention towards Multi Instance Learning (MIL) algorithms where labels are available at a bag level rather than at an instance level. In this chapter, we motivate the need for MIL algorithms and describe an ensemble based method, wherein the members of the ensemble are lazy learning classifiers using the Citation Nearest Neighbour method. Diversity among the ensemble methods is achieved by optimizing their parameters using a multi-objective optimization method, with the objective being to maximize positive class accuracy and minimize false positive rate. We demonstrate results of the methodology on the standard Musk 1 dataset.

scholarly journals Spelling Correction for Text Documents in Bahasa Indonesia Using Finite State Automata and Levinshtein Distance Method

2018 ◽  
Vol 164 ◽  
pp. 01047 ◽  
Author(s):  
Viny Christanti Mawardi ◽  
Niko Susanto ◽  
Dali Santun Naga
Keyword(s):  
False Positive ◽  
False Positive Rate ◽  
Training Data ◽  
Levenshtein Distance ◽  
Finite State Automata ◽  
Distance Method ◽  
Spelling Correction ◽  
Hit Rate ◽  
Finite State ◽  
Positive Rate

Any mistake in writing of a document will cause the information to be told falsely. These days, most of the document is written with a computer. For that reason, spelling correction is needed to solve any writing mistakes. This design process discuss about the making of spelling correction for document text in Indonesian language with document's text as its input and a .txt file as its output. For the realization, 5 000 news articles have been used as training data. Methods used includes Finite State Automata (FSA), Levenshtein distance, and N-gram. The results of this designing process are shown by perplexity evaluation, correction hit rate and false positive rate. Perplexity with the smallest value is a unigram with value 1.14. On the other hand, the highest percentage of correction hit rate is bigram and trigram with value 71.20 %, but bigram is superior in processing time average which is 01:21.23 min. The false positive rate of unigram, bigram, and trigram has the same percentage which is 4.15 %. Due to the disadvantages at using FSA method, modification is done and produce bigram's correction hit rate as high as 85.44 %.

scholarly journals Detecting Website Defacements Based on Machine Learning Techniques and Attack Signatures

Computers ◽  
2019 ◽  
Vol 8 (2) ◽  
pp. 35 ◽  
Author(s):  
Xuan Dau Hoang ◽  
Ngoc Tuong Nguyen
Keyword(s):  
Machine Learning ◽  
Web Applications ◽  
False Positive Rate ◽  
Training Data ◽  
Machine Learning Techniques ◽  
Web Pages ◽  
Government Organizations ◽  
Detection Model ◽  
Learning Techniques ◽  
Positive Rate

Defacement attacks have long been considered one of prime threats to websites and web applications of companies, enterprises, and government organizations. Defacement attacks can bring serious consequences to owners of websites, including immediate interruption of website operations and damage of the owner reputation, which may result in huge financial losses. Many solutions have been researched and deployed for monitoring and detection of website defacement attacks, such as those based on checksum comparison, diff comparison, DOM tree analysis, and complicated algorithms. However, some solutions only work on static websites and others demand extensive computing resources. This paper proposes a hybrid defacement detection model based on the combination of the machine learning-based detection and the signature-based detection. The machine learning-based detection first constructs a detection profile using training data of both normal and defaced web pages. Then, it uses the profile to classify monitored web pages into either normal or attacked. The machine learning-based component can effectively detect defacements for both static pages and dynamic pages. On the other hand, the signature-based detection is used to boost the model’s processing performance for common types of defacements. Extensive experiments show that our model produces an overall accuracy of more than 99.26% and a false positive rate of about 0.27%. Moreover, our model is suitable for implementation of a real-time website defacement monitoring system because it does not demand extensive computing resources.

scholarly journals Using Logistic Regression to Estimate the False Positive Rate in The IDI (SoLinks)

2020 ◽  
Vol 5 (5) ◽  
Author(s):  
Anna Lin ◽  
Soon Song ◽  
Nancy Wang
Keyword(s):  
Logistic Regression ◽  
False Positive ◽  
False Positive Rate ◽  
Training Data ◽  
Model Parameters ◽  
Data Infrastructure ◽  
True Match ◽  
Positive Rate ◽  
Research Populations ◽  
The Individual

IntroductionStats NZ’s Integrated Data Infrastructure (IDI) is a linked longitudinal database combining administrative and survey data. Previously, false positive linkages (FP) in the IDI were assessed by clerical review of a sample of linked records, which was time consuming and subject to inconsistency. Objectives and ApproachA modelled approach, ‘SoLinks’ has been developed in order to automate the FP estimation process for the IDI. It uses a logistic regression model to calculate the probability that a given link is a true match. The model is based on the agreement types defined for four key linking variables – first name, last name, sex, and date of birth. Exemptions have been given to some specific types of links that we believe to be high quality true matches. The training data used to estimate the model parameters was based on the outcomes of the clerical review process over several years. ResultsWe have compared the FP rates estimated through clerical review to the ones estimated through the SoLinks model. Some SoLinks estimates fall outside the 95% confidence intervals of the clerically reviewed ones. This may be the result of the pre-defined probabilities for the specific types of links are too high. ConclusionThe automation of FP checking has saved analyst time and resource. The modelled FP estimates have been more stable across time than the previous clerical reviews. As this model estimates the probability of a true match at the individual link level, we may provide this probability to researchers so that they can calculate linked quality indicators for their research populations.

scholarly journals Var-CNN: A Data-Efficient Website Fingerprinting Attack Based on Deep Learning

2019 ◽  
Vol 2019 (4) ◽  
pp. 292-310 ◽  
Author(s):  
Sanjit Bhat ◽  
David Lu ◽  
Albert Kwon ◽  
Srinivas Devadas
Keyword(s):  
Deep Learning ◽  
State Of The Art ◽  
False Positive Rate ◽  
True Positive Rate ◽  
Training Data ◽  
Open World ◽  
Prior Art ◽  
Lower False Positive Rate ◽  
Positive Rate ◽  
Fingerprinting Attack

Abstract In recent years, there have been several works that use website fingerprinting techniques to enable a local adversary to determine which website a Tor user visits. While the current state-of-the-art attack, which uses deep learning, outperforms prior art with medium to large amounts of data, it attains marginal to no accuracy improvements when both use small amounts of training data. In this work, we propose Var-CNN, a website fingerprinting attack that leverages deep learning techniques along with novel insights specific to packet sequence classification. In open-world settings with large amounts of data, Var-CNN attains over 1% higher true positive rate (TPR) than state-of-the-art attacks while achieving 4× lower false positive rate (FPR). Var-CNN’s improvements are especially notable in low-data scenarios, where it reduces the FPR of prior art by 3.12% while increasing the TPR by 13%. Overall, insights used to develop Var-CNN can be applied to future deep learning based attacks, and substantially reduce the amount of training data needed to perform a successful website fingerprinting attack. This shortens the time needed for data collection and lowers the likelihood of having data staleness issues.

scholarly journals Overcoming the limitations of patch-based learning to detect cancer in whole slide images

2021 ◽  
Vol 11 (1) ◽  
Author(s):  
Ozan Ciga ◽  
Tony Xu ◽  
Sharon Nofech-Mozes ◽  
Shawna Noy ◽  
Fang-I Lu ◽  
...  
Keyword(s):  
Multiple Scales ◽  
False Positive Rate ◽  
Sampling Strategy ◽  
Image Features ◽  
Training Data ◽  
Final Decision ◽  
Data Sampling ◽  
Negative Data ◽  
Image Patches ◽  
Whole Slide Images

AbstractWhole slide images (WSIs) pose unique challenges when training deep learning models. They are very large which makes it necessary to break each image down into smaller patches for analysis, image features have to be extracted at multiple scales in order to capture both detail and context, and extreme class imbalances may exist. Significant progress has been made in the analysis of these images, thanks largely due to the availability of public annotated datasets. We postulate, however, that even if a method scores well on a challenge task, this success may not translate to good performance in a more clinically relevant workflow. Many datasets consist of image patches which may suffer from data curation bias; other datasets are only labelled at the whole slide level and the lack of annotations across an image may mask erroneous local predictions so long as the final decision is correct. In this paper, we outline the differences between patch or slide-level classification versus methods that need to localize or segment cancer accurately across the whole slide, and we experimentally verify that best practices differ in both cases. We apply a binary cancer detection network on post neoadjuvant therapy breast cancer WSIs to find the tumor bed outlining the extent of cancer, a task which requires sensitivity and precision across the whole slide. We extensively study multiple design choices and their effects on the outcome, including architectures and augmentations. We propose a negative data sampling strategy, which drastically reduces the false positive rate (25% of false positives versus 62.5%) and improves each metric pertinent to our problem, with a 53% reduction in the error of tumor extent. Our results indicate classification performances of image patches versus WSIs are inversely related when the same negative data sampling strategy is used. Specifically, injection of negatives into training data for image patch classification degrades the performance, whereas the performance is improved for slide and pixel-level WSI classification tasks. Furthermore, we find applying extensive augmentations helps more in WSI-based tasks compared to patch-level image classification.

scholarly journals Intelligent Frame Work to Predict Autism in Infants using Machine Learning

2020 ◽  
Vol 9 (1) ◽  
pp. 675-679
Keyword(s):  
Machine Learning ◽  
Cognitive Abilities ◽  
False Positive Rate ◽  
Autism Spectrum ◽  
Spectrum Disorder ◽  
Training Data ◽  
Screening Tests ◽  
Behavioral Impairment ◽  
Medical Checkup ◽  
Positive Rate

Autistic Spectrum Disorder (ASD) is a behavioral impairment that interferes with the use of auditory, communicative, cognitive, abilities and social skills. ADS was introduce researched using advanced approaches based on machine learning to speed up the diagnostic period or increase the responsiveness, reliability or precision of the diagnosis process. Normal medical checkup data (training data) of the baby, test data is used to classify the child with autism. The assessment results showed that, for both types of datasets, the proposed prediction model produces better results in terms of precision, responsiveness, precision and false positive rate (FPR). In the modern day, Autism Spectrum Disorder (ASD) is gaining some momentum quicker than ever. Detecting signs of autism by screening tests is very costly and time consuming. Autism can be identified faster by combining artificial intelligence and Machine Learning (ML)

scholarly journals Evaluating GAN-Based Image Augmentation for Threat Detection in Large-Scale Xray Security Images

2020 ◽  
Vol 11 (1) ◽  
pp. 36
Author(s):  
Joanna Kazzandra Dumagpi ◽  
Yong-Jin Jeong
Keyword(s):  
Computer Vision ◽  
Large Scale ◽  
False Positive Rate ◽  
Image Synthesis ◽  
True Positive Rate ◽  
Generative Adversarial Networks ◽  
X Ray ◽  
Imbalance Problem ◽  
High True Positive Rate ◽  
Positive Rate

The inherent imbalance in the data distribution of X-ray security images is one of the most challenging aspects of computer vision algorithms applied in this domain. Most of the prior studies in this field have ignored this aspect, limiting their application in the practical setting. This paper investigates the effect of employing Generative Adversarial Networks (GAN)-based image augmentation, or image synthesis, in improving the performance of computer vision algorithms on an imbalanced X-ray dataset. We used Deep Convolutional GAN (DCGAN) to generate new X-ray images of threat objects and Cycle-GAN to translate camera images of threat objects to X-ray images. We synthesized new X-ray security images by combining threat objects with background X-ray images, which are used to augment the dataset. Then, we trained various Faster (Region Based Convolutional Neural Network) R-CNN models using different augmentation approaches and evaluated their performance on a large-scale practical X-ray image dataset. Experiment results show that image synthesis is an effective approach to combating the imbalance problem by significantly reducing the false-positive rate (FPR) by up to 15.3%. The FPR is further improved by up to 19.9% by combining image synthesis and conventional image augmentation. Meanwhile, a relatively high true positive rate (TPR) of about 94% was maintained regardless of the augmentation method used.

Sign in / Sign up

Export Citation Format

Share Document