Search results for: classification methods

Authors: Mohammed Abdullah Alshahrani

Abstract:

The research focuses on investigating the use of partial least squares (PLS) methodology for addressing challenges associated with high-dimensional correlated data. Recent technological advancements have led to experiments producing data characterized by a large number of variables compared to observations, with substantial inter-variable correlations. Such data patterns are common in chemometrics, where near-infrared (NIR) spectrometer calibrations record chemical absorbance levels across hundreds of wavelengths, and in genomics, where thousands of genomic regions' copy number alterations (CNA) are recorded from cancer patients. PLS serves as a widely used method for analyzing high-dimensional data, functioning as a regression tool in chemometrics and a classification method in genomics. It handles data complexity by creating latent variables (components) from original variables. However, applying PLS can present challenges. The study investigates key areas to address these challenges, including unifying interpretations across three main PLS algorithms and exploring unusual negative shrinkage factors encountered during model fitting. The research presents an alternative approach to addressing the interpretation challenge of predictor weights associated with PLS. Sparse estimation of predictor weights is employed using a penalty function combining a lasso penalty for sparsity and a Cauchy distribution-based penalty to account for variable dependencies. The results demonstrate sparse and grouped weight estimates, aiding interpretation and prediction tasks in genomic data analysis. High-dimensional data scenarios, where predictors outnumber observations, are common in regression analysis applications. Ordinary least squares regression (OLS), the standard method, performs inadequately with high-dimensional and highly correlated data. Copy number alterations (CNA) in key genes have been linked to disease phenotypes, highlighting the importance of accurate classification of gene expression data in bioinformatics and biology using regularized methods like PLS for regression and classification.

Keywords: partial least square regression, genetics data, negative filter factors, high dimensional data, high correlated data

Procedia PDF Downloads 44

Authors: H. G. C. P. Dinesh, G. Tharshini, M. P. B. Ekanayake, G. M. R. I. Godaliyadda

Abstract:

This paper presents a new spatial feature extraction method based on principle component analysis (PCA) and Fisher Discernment Analysis (FDA) for facial expression recognition. It not only extracts reliable features for classification, but also reduces the feature space dimensions of pattern samples. In this method, first each gray scale image is considered in its entirety as the measurement matrix. Then, principle components (PCs) of row vectors of this matrix and variance of these row vectors along PCs are estimated. Therefore, this method would ensure the preservation of spatial information of the facial image. Afterwards, by incorporating the spectral information of the eigen-filters derived from the PCs, a feature vector was constructed, for a given image. Finally, FDA was used to define a set of basis in a reduced dimension subspace such that the optimal clustering is achieved. The method of FDA defines an inter-class scatter matrix and intra-class scatter matrix to enhance the compactness of each cluster while maximizing the distance between cluster marginal points. In order to matching the test image with the training set, a cosine similarity based Bayesian classification was used. The proposed method was tested on the Cohn-Kanade database and JAFFE database. It was observed that the proposed method which incorporates spatial information to construct an optimal feature space outperforms the standard PCA and FDA based methods.

Keywords: facial expression recognition, principle component analysis (PCA), fisher discernment analysis (FDA), eigen-filter, cosine similarity, bayesian classifier, f-measure

Procedia PDF Downloads 422

Authors: Rosa Figueroa, Christopher Flores

Abstract:

Text categorization is the problem of assigning a new document to a set of predetermined categories, on the basis of a training set of free-text data that contains documents whose category membership is known. To train a classification model, it is necessary to extract characteristics in the form of tokens that facilitate the learning and classification process. In text categorization, the feature extraction process involves the use of word sequences also known as N-grams. In general, it is expected that documents belonging to the same category share similar features. The Smith-Waterman (SW) algorithm is a dynamic programming algorithm that performs a local sequence alignment in order to determine similar regions between two strings or protein sequences. This work explores the use of SW algorithm as an alternative to feature extraction in text categorization. The dataset used for this purpose, contains 2,610 annotated documents with the classes Obese/Non-Obese. This dataset was represented in a matrix form using the Bag of Word approach. The score selected to represent the occurrence of the tokens in each document was the term frequency-inverse document frequency (TF-IDF). In order to extract features for classification, four experiments were conducted: the first experiment used SW to extract features, the second one used unigrams (single word), the third one used bigrams (two word sequence) and the last experiment used a combination of unigrams and bigrams to extract features for classification. To test the effectiveness of the extracted feature set for the four experiments, a Support Vector Machine (SVM) classifier was tuned using 20% of the dataset. The remaining 80% of the dataset together with 5-Fold Cross Validation were used to evaluate and compare the performance of the four experiments of feature extraction. Results from the tuning process suggest that SW performs better than the N-gram based feature extraction. These results were confirmed by using the remaining 80% of the dataset, where SW performed the best (accuracy = 97.10%, weighted average F-measure = 97.07%). The second best was obtained by the combination of unigrams-bigrams (accuracy = 96.04, weighted average F-measure = 95.97) closely followed by the bigrams (accuracy = 94.56%, weighted average F-measure = 94.46%) and finally unigrams (accuracy = 92.96%, weighted average F-measure = 92.90%).

Keywords: comorbidities, machine learning, obesity, Smith-Waterman algorithm

Procedia PDF Downloads 290

Authors: Ceren Kaya, Okan Erkaymaz, Orhan Ayar, Mahmut Özer

Abstract:

Diabetes Mellitus (Diabetes) is a disease based on insulin hormone disorders and causes high blood glucose. Clinical findings determine that diabetes can be diagnosed by electrophysiological signals obtained from the vital organs. 'Diabetic Retinopathy' is one of the most common eye diseases resulting on diabetes and it is the leading cause of vision loss due to structural alteration of the retinal layer vessels. In this study, features of horizontal and vertical Video-Oculography (VOG) signals have been used to classify non-proliferative and proliferative diabetic retinopathy disease. Twenty-five features are acquired by using discrete wavelet transform with VOG signals which are taken from 21 subjects. Two models, based on multi-layer perceptron and radial basis function, are recommended in the diagnosis of Diabetic Retinopathy. The proposed models also can detect level of the disease. We show comparative classification performance of the proposed models. Our results show that proposed the RBF model (100%) results in better classification performance than the MLP model (94%).

Keywords: diabetic retinopathy, discrete wavelet transform, multi-layer perceptron, radial basis function, video-oculography (VOG)

Procedia PDF Downloads 251

Authors: Yoshimoto Kurihara, Tad Gonsalves

Abstract:

Currently, in the field of object posture estimation, there is research on estimating the position and angle of an object by storing a 3D model of the object to be estimated in advance in a computer and matching it with the model. However, in this research, we have succeeded in creating a module that is much simpler, smaller in scale, and faster in operation. Our 6D pose estimation model consists of two different networks – a classification network and a regression network. From a single RGB image, the trained model estimates the class of the object in the image, the coordinates of the object, and its rotation angle in 3D space. In addition, we compared the estimation accuracy of each camera position, i.e., the angle from which the object was captured. The highest accuracy was recorded when the camera position was 75°, the accuracy of the classification was about 87.3%, and that of regression was about 98.9%.

Keywords: 6D posture estimation, image recognition, deep learning, AlexNet

Procedia PDF Downloads 148

Authors: Xiaoqi Jia, Qing Zhu, Hao Zhang, Su Yang

Abstract:

A gender recognition system is able to tell the gender of the given person through a few of frontal facial images. An effective gender recognition approach enables to improve the performance of many other applications, including security monitoring, human-computer interaction, image or video retrieval and so on. In this paper, we present an effective method for gender classification task in frontal facial images based on deep belief networks (DBNs), which can pre-train model and improve accuracy a little bit. Our experiments have shown that the pre-training method with DBNs for gender classification task is feasible and achieves a little improvement of accuracy on FERET and CAS-PEAL-R1 facial datasets.

Keywords: gender recognition, beep belief net-works, semi-supervised learning, greedy-layer wise RBMs

Procedia PDF Downloads 444

Authors: Oumaima Khlifati, Khadija Baba

Abstract:

Pavement distress is the main factor responsible for the deterioration of road structure durability, damage vehicles, and driver comfort. Transportation agencies spend a high proportion of their funds on pavement monitoring and maintenance. The auscultation of pavement distress was based on the manual survey, which was extremely time consuming, labor intensive, and required domain expertise. Therefore, the automatic distress detection is needed to reduce the cost of manual inspection and avoid more serious damage by implementing the appropriate remediation actions at the right time. Inspired by recent deep learning applications, this paper proposes an algorithm for automatic road distress detection and classification using on the Deep Convolutional Neural Network (DCNN). In this study, the types of pavement distress are classified as transverse or longitudinal cracking, alligator, pothole, and intact pavement. The dataset used in this work is composed of public asphalt pavement images. In order to learn the structure of the different type of distress, the DCNN models are trained and tested as a multi-label classification task. In addition, to get the highest accuracy for our model, we adjust the structural optimization hyper parameters such as the number of convolutions and max pooling, filers, size of filters, loss functions, activation functions, and optimizer and fine-tuning hyper parameters that conclude batch size and learning rate. The optimization of the model is executed by checking all feasible combinations and selecting the best performing one. The model, after being optimized, performance metrics is calculated, which describe the training and validation accuracies, precision, recall, and F1 score.

Keywords: distress pavement, hyperparameters, automatic classification, deep learning

Procedia PDF Downloads 82

Authors: Dimitar Karastoyanov, Todor Penchev

Abstract:

The paper presents the research of a kind of autonomous mobile robots, intended for work and adaptive perception in unknown and unstructured environment. The objective are robots, dedicated for multi-sensory environment perception and exploration, like measurements and samples taking, discovering and putting a mark on the objects as well as environment interactions–transportation, carrying in and out of equipment and objects. At that ground classification of the different types mobile robots in accordance with the way of locomotion (wheel- or chain-driven, walking, etc.), used drive mechanisms, kind of sensors, end effectors, area of application, etc. is made. Modular system for the mechanical construction of the mobile robots is proposed. Special PLC on the base of AtMega128 processor for robot control is developed. Electronic modules for the wireless communication on the base of Jennic processor as well as the specific software are developed. The methods, means and algorithms for adaptive environment behaviour and tasks realization are examined. The methods of group control of mobile robots and for suspicious objects detecting and handling are discussed too.

Keywords: mobile robots, wireless communications, environment investigations, group control, suspicious objects

Procedia PDF Downloads 346

Authors: Fuad M. Alkoot

Abstract:

The main component of a machine learning system is the classifier. Classifiers are mathematical models that can perform classification tasks for a specific application area. Additionally, many classifiers are combined using any of the available methods to reduce the classifier error rate. The benefits gained from the combination of multiple classifier designs has motivated the development of diverse approaches to multiple classifiers. We aim to investigate using fractal geometry to develop an improved classifier combiner. Initially we experiment with measuring the fractal dimension of data and use the results in the development of a combiner strategy.

Keywords: fractal geometry, machine learning, classifier, fractal dimension

Procedia PDF Downloads 206

Authors: Thrivikraman Aswathi, S. Advaith

Abstract:

As there can be cases where the use of real data is somehow limited, such as when it is hard to get access to a large volume of real data, we need to go for synthetic data generation. This produces high-quality synthetic data while maintaining the statistical properties of a specific dataset. In the present work, a generative adversarial network (GAN) is trained to produce multivariate time series (MTS) data since the MTS is now being gathered more often in various real-world systems. Furthermore, the GAN-generated MTS data is fed into a transformer-based deep learning architecture that carries out the data categorization into predefined classes. Further, the model is evaluated across various distinct domains by generating corresponding MTS data.

Keywords: GAN, transformer, classification, multivariate time series

Procedia PDF Downloads 123

Authors: Muthukumarasamy Govindarajan

Abstract:

Detection of unwanted, unsolicited mails called spam from email is an interesting area of research. It is necessary to evaluate the performance of any new spam classifier using standard data sets. Recently, ensemble-based classifiers have gained popularity in this domain. In this research work, an efficient email filtering approach based on ensemble methods is addressed for developing an accurate and sensitive spam classifier. The proposed approach employs Naive Bayes (NB), Support Vector Machine (SVM) and Genetic Algorithm (GA) as base classifiers along with different ensemble methods. The experimental results show that the ensemble classifier was performing with accuracy greater than individual classifiers, and also hybrid model results are found to be better than the combined models for the e-mail dataset. The proposed ensemble-based classifiers turn out to be good in terms of classification accuracy, which is considered to be an important criterion for building a robust spam classifier.

Keywords: accuracy, arcing, bagging, genetic algorithm, Naive Bayes, sentiment mining, support vector machine

Procedia PDF Downloads 135

Authors: Subhadeep Mandal, Sujoy Bhattacharya, Pabitra Mitra, Diya Guha Roy, Seema Bhattacharya

Abstract:

E-commerce firms allow customers to evaluate and review the things they buy as a positive or bad experience. The e-commerce transaction processes are made up of a variety of diverse organizations and activities that operate independently but are connected together to complete the transaction (from placing an order to the goods reaching the client). After a negative shopping experience, clients frequently disregard the critical assessment of these businesses and submit their feedback on an all-over basis, which benefits certain enterprises but is tedious for others. In this article, we solely dealt with negative reviews and attempted to distinguish between negative reviews where the e-commerce firm is explicitly blamed by customers for a bad purchasing experience and other negative reviews.

Keywords: e-commerce, online shopping, customer reviews, customer behaviour, text analytics, n-grams classification

Procedia PDF Downloads 253

Authors: Neda Orak, Mostafa Zarei

Abstract:

Environmental assessment is an important session in environment management. Since various methods and techniques have been produces and implemented. Remote sensing (RS) is widely used in many scientific and research fields such as geology, cartography, geography, agriculture, forestry, land use planning, environment, etc. It can show earth surface objects cyclical changes. Also, it can show earth phenomena limits on basis of electromagnetic reflectance changes and deviations records. The research has been done on mangrove forests assessment by RS techniques. Mangrove forests quantitative analysis in Basatin and Bidkhoon estuaries was the aim of this research. It has been done by Landsat satellite images from 1975- 2013 and match to ground control points. This part of mangroves are the last distribution in northern hemisphere. It can provide a good background to improve better management on this important ecosystem. Landsat has provided valuable images to earth changes detection to researchers. This research has used MSS, TM, +ETM, OLI sensors from 1975, 1990, 2000, 2003-2013. Changes had been studied after essential corrections such as fix errors, bands combination, georeferencing on 2012 images as basic image, by maximum likelihood and IPVI Index. It was done by supervised classification. 2004 google earth image and ground points by GPS (2010-2012) was used to compare satellite images obtained changes. Results showed mangrove area in bidkhoon was 1119072 m2 by GPS and 1231200 m2 by maximum likelihood supervised classification and 1317600 m2 by IPVI in 2012. Basatin areas is respectively: 466644 m2, 88200 m2, 63000 m2. Final results show forests have been declined naturally. It is due to human activities in Basatin. The defect was offset by planting in many years. Although the trend has been declining in recent years again. So, it mentioned satellite images have high ability to estimation all environmental processes. This research showed high correlation between images and indexes such as IPVI and NDVI with ground control points.

Keywords: IPVI index, Landsat sensor, maximum likelihood supervised classification, Nayband National Park

Procedia PDF Downloads 286

Authors: Jamshid Qodirov, Sirojiddin Komolov, Ravilov Mirahmad, Olimjon Mirzayev

Abstract:

Nowadays, natural language processing tools play a crucial role in our daily lives, including various techniques with text processing. There are very advanced models in modern languages, such as English, Russian etc. But, in some languages, such as Uzbek, the NLP models have been developed recently. Thus, there are only a few NLP models in Uzbek language. Moreover, there is no such work that could show which Uzbek NLP model behaves in different situations and when to use them. This work tries to close this gap and compares the Uzbek NLP models existing as of the time this article was written. The authors try to compare the NLP models in two different scenarios: sentiment analysis and sentence similarity, which are the implementations of the two most common problems in the industry: classification and similarity. Another outcome from this work is two datasets for classification and sentence similarity in Uzbek language that we generated ourselves and can be useful in both industry and academia as well.

Keywords: NLP, benchmak, bert, vectorization

Procedia PDF Downloads 49

Authors: Carolyn Whyte, Tina Halai, Sonita Koshal

Abstract:

Aim: There are many methods available to assess the potential difficulty of third molar surgery. This study investigated various factors to assess whether they had a bearing on the difficulties encountered. Study design: A retrospective study was completed of 62 single mandibular third molar teeth removed under day case general anaesthesia between May 2016 and August 2016 by 3 consultant oral surgeons. Method: Data collection was by examining the OPG radiographs of each tooth and recording the necessary data. This was depth of impaction, angulation, bony impaction, point of application in relation to second molar, root morphology, Pell and Gregory classification and Winters Lines. This was completed by one assessor and verified by another. Information on medical history, anxiety, ethnicity and age were recorded. Case notes and surgical entries were examined for any difficulties encountered. Results: There were 5 cases which encountered surgical difficulties which included fracture of root apices (3) which were left in situ, prolonged bleeding (1) and post-operative numbness >6 months(1). Four of the 5 cases had Pell and Gregory classification as (B) where the occlusal plane of the impacted tooth is between the occlusal plane and the cervical line of the adjacent tooth. 80% of cases had the point of application as either coronal or apical one third (1/3) in relation to the second molar. However, there was variability in all other aspects of assessment in predicting difficulty of removal. Conclusions: Of the cases which encountered difficulties they all had at least one predictor of potential complexity but these varied case by case.

Keywords: impaction, mandibular third molar, radiographic assessment, surgical removal

Procedia PDF Downloads 178

Authors: Natthawut Arin

Abstract:

In forensic contexts, competency to stand trial assessments are the most common referrals. The defendants may attempt to endorse psychopathology symptoms and feign incompetent. Coaching, which can be teaching them test-taking strategies to avoid detection of psychopathological symptoms feigning. Recently, the Symptom Validity Testings (SVTs) were created to detect feigning. Moreover, the works of the literature showed that the effects of coaching on SVTs may be more robust to the effects of coaching. Thai Symptom Validity Test (SVT-Th) was designed as SVTs which demonstrated adequate psychometric properties and ability to classify between feigners and honest responders. Thus, the current study to examine the utility as the robustness of SVT-Th in the detection of feigned psychopathology. Participants consisted of 120 were recruited from undergraduate courses in psychology, randomly assigned to one of three groups. The SVT-Th was administered to those three scenario-experimental groups: (a) Uncoached group were asked to respond honestly (n=40), (b) Symptom-coached without warning group were asked to feign psychiatric symptoms to gain incompetency to stand trial (n=40), while (c) Test-coached with warning group were asked to feign psychiatric symptoms to avoid test detection but being incompetency to stand trial (n=40). Group differences were analyzed using one-way ANOVAs. The result revealed an uncoached group (M = 4.23, SD.= 5.20) had significantly lower SVT-Th mean scores than those both coached groups (M =185.00, SD.= 72.88 and M = 132.10, SD.= 54.06, respectively). Classification rates were calculated to determine the classification accuracy. Result indicated that SVT-Th had overall classification accuracy rates of 96.67% with acceptable of 95% sensitivity and 100% specificity rates. Overall, the results of the present study indicate that the SVT-Th yielded high adequate indices of accuracy and these findings suggest that the SVT-Th is robustness against coaching.

Keywords: incompetency to stand trial, coaching, robustness, classification accuracy

Procedia PDF Downloads 126

Authors: Ahlem Mougaida, Hedi Bel Hadj Salah

Abstract:

Several methods are suggested to improve the numerical integration in Galerkin weak form for Meshless methods. In fact, integrating without taking into account the characteristics of the shape functions reproduced by Meshless methods (rational functions, with compact support etc.), causes a large integration error that influences the PDE’s approximate solution. Comparisons between different methods of numerical integration for rational functions are discussed and compared. The algorithms are implemented in Matlab. Finally, numerical results were presented to prove the efficiency of our algorithms in improving results.

Keywords: adaptive methods, meshless, numerical integration, rational quadrature

Procedia PDF Downloads 354

Authors: Levent Basayigit, Sinan Demir, Yusuf Ucar, Burhan Kara

Abstract:

Many research projects require accurate delineation of the different land cover type of the agricultural area. Especially it is critically important for the definition of specific plants like cannabis. However, the complexity of vegetation stands structure, abundant vegetation species, and the smooth transition between different seconder section stages make vegetation classification difficult when using traditional approaches such as the maximum likelihood classifier. Most of the time, classification distinguishes only between trees/annual or grain. It has been difficult to accurately determine the cannabis mixed with other plants. In this paper, a mixed distribution models approach is applied to classify pure and mix cannabis parcels using Worldview-2 imagery in the Lakes region of Turkey. Five different land use types (i.e. sunflower, maize, bare soil, and cannabis) were identified in the image. A constrained Gaussian mixture discriminant analysis (GMDA) was used to unmix the image. In the study, 255 reflectance ratios derived from spectral signatures of seven bands (Blue-Green-Yellow-Red-Rededge-NIR1-NIR2) were randomly arranged as 80% for training and 20% for test data. Gaussian mixed distribution model approach is proved to be an effective and convenient way to combine very high spatial resolution imagery for distinguishing cannabis vegetation. Based on the overall accuracies of the classification, the Gaussian mixed distribution model was found to be very successful to achieve image classification tasks. This approach is sensitive to capture the illegal cannabis planting areas in the large plain. This approach can also be used for monitoring and determination with spectral reflections in illegal cannabis planting areas.

Keywords: Gaussian mixture discriminant analysis, spectral mixture model, Worldview-2, land parcels

Procedia PDF Downloads 191

Authors: Chelly Souhir, Harizi Chahida, Hachaichi Aicha, Aissaoui Sihem, Chahed Mohamed Kouni

Abstract:

Introduction: In Tunisia, few studies have attempted to describe the demand for primary care in a standardized and systematic way. The purpose of this study is to describe the main reasons for demand for care in primary health care, through a survey of the Ariana Governorate PHC and to identify their evolutionary trend compared to last 30 years, reported by studies of the same type. Materials and methods: This is a cross-sectional descriptive study which concerns the study of consultants in the first line of the governorate of Ariana and their use of care recorded during 2 days in the same week during the month of May 2016, in each of these PHC. The same data collection sheet was used in all CSBs. The coding of the information was done according to the International Classification of Primary Care (ICPC). The data was entered and analyzed by the EPI Info 7 software. Results: Our study found that the most common ICPC chapters are respiratory (42%) and digestive (13.2%). In 1996 were the respiratory (43.5%) and circulatory (7.8%). In 2000, we found also the respiratory (39,6%) and circulatory (10,9%). In 2002, respiratory (43%) and digestive (10.1%) motives were the most frequent. According to the ICPC, the pathologies in our study were acute angina (19%), acute bronchitis and bronchiolitis (8%). In 1996, it was tonsillitis ( 21.6%) and acute bronchitis (7.2%). For Ben Abdelaziz in 2000, tonsillitis (14.5%) follow by acute bronchitis (8.3%). In 2002, acute angina (15.7%), acute bronchitis and bronchiolitis (11.2%) were the most common. Conclusion: Acute angina and tonsillitis are the most common in all studies conducted in Tunisia.

Keywords: acute angina, classification of primary care, primary health care, tonsillitis, Tunisia

Procedia PDF Downloads 519

Authors: Kyle Saltmarsh

Abstract:

Acoustic testing possesses many benefits due to its non-destructive nature and practicality. There hence exists many scenarios in which using acoustic testing for condition testing shows powerful feasibility. A wealth of information is contained within the acoustic and vibration characteristics of structures, allowing the development meaningful features for the classification of their respective condition. In this paper, methods, results, and discussions are presented on the use of non-destructive acoustic testing coupled with acoustic feature extraction and machine learning techniques for the condition testing of manufactured circular steel plates subjected to varied levels of damage.

Keywords: plates, deformation, acoustic features, machine learning

Procedia PDF Downloads 330

Authors: Jonathan Sohn

Abstract:

Deep neural networks (DNNs) have shown state-of-the-art performance for many applications, including computer vision, natural language processing, and speech recognition. Recently, adversarial attacks have been studied in the context of deep neural networks, which aim to alter the results of deep neural networks by modifying the inputs slightly. For example, an adversarial attack on a DNN used for object detection can cause the DNN to miss certain objects. As a result, the reliability of DNNs is undermined by their lack of robustness against adversarial attacks, raising concerns about their use in safety-critical applications such as autonomous driving. In this paper, we focus on studying the adversarial attacks and defenses on DNNs for image classification. There are two types of adversarial attacks studied which are fast gradient sign method (FGSM) attack and projected gradient descent (PGD) attack. A DNN forms decision boundaries that separate the input images into different categories. The adversarial attack slightly alters the image to move over the decision boundary, causing the DNN to misclassify the image. FGSM attack obtains the gradient with respect to the image and updates the image once based on the gradients to cross the decision boundary. PGD attack, instead of taking one big step, repeatedly modifies the input image with multiple small steps. There is also another type of attack called the target attack. This adversarial attack is designed to make the machine classify an image to a class chosen by the attacker. We can defend against adversarial attacks by incorporating adversarial examples in training. Specifically, instead of training the neural network with clean examples, we can explicitly let the neural network learn from the adversarial examples. In our experiments, the digit recognition accuracy on the MNIST dataset drops from 97.81% to 39.50% and 34.01% when the DNN is attacked by FGSM and PGD attacks, respectively. If we utilize FGSM training as a defense method, the classification accuracy greatly improves from 39.50% to 92.31% for FGSM attacks and from 34.01% to 75.63% for PGD attacks. To further improve the classification accuracy under adversarial attacks, we can also use a stronger PGD training method. PGD training improves the accuracy by 2.7% under FGSM attacks and 18.4% under PGD attacks over FGSM training. It is worth mentioning that both FGSM and PGD training do not affect the accuracy of clean images. In summary, we find that PGD attacks can greatly degrade the performance of DNNs, and PGD training is a very effective way to defend against such attacks. PGD attacks and defence are overall significantly more effective than FGSM methods.

Keywords: deep neural network, adversarial attack, adversarial defense, adversarial machine learning

Procedia PDF Downloads 189

Authors: Hyung Jun Ahn

Abstract:

With the growing importance of social media, it is imperative to analyze it to understand the users. Users share useful information and their experience through social media, where much of what is shared is in the form of texts. This study focused on blogs and aimed to test whether the psycho-linguistic characteristics of blog texts vary with the subject or the type of experience of the texts. For this goal, blog texts about four different types of experience, Go, skiing, reading, and musical were collected through the search API of the Tistory blog service. The analysis of the texts showed that various psycholinguistic characteristics of the texts are different across the four categories of the texts. Moreover, the machine learning experiment using the characteristics for automatic text classification showed significant performance. Specifically, the ensemble method, based on functional tree and bagging appeared to be most effective in classification.

Keywords: blog, social media, text analysis, psycholinguistics

Procedia PDF Downloads 275

Authors: Rujia Chen, Ajit Narayanan

Abstract:

Convolutional neural networks (CNN), as typically applied in deep learning, use layer-wise backpropagation (BP) to construct filters and kernels for feature extraction. Such filters are 2D or 3D groups of weights for constructing feature maps at subsequent layers of the CNN and are shared across the entire input. BP as a gradient descent algorithm has well-known problems of getting stuck at local optima. The use of genetic algorithms (GAs) for evolving weights between layers of standard artificial neural networks (ANNs) is a well-established area of neuroevolution. In particular, the use of crossover techniques when optimizing weights can help to overcome problems of local optima. However, the application of GAs for evolving the weights of filters and kernels in CNNs is not yet an established area of neuroevolution. In this paper, a GA-based filter development algorithm is proposed. The results of the proof-of-concept experiments described in this paper show the proposed GA algorithm can find filter weights through evolutionary techniques rather than BP learning. For some simple classification tasks like geometric shape recognition, the proposed algorithm can achieve 100% accuracy. The results for MNIST classification, while not as good as possible through standard filter learning through BP, show that filter and kernel evolution warrants further investigation as a new subarea of neuroevolution for deep architectures.

Keywords: neuroevolution, convolutional neural network, genetic algorithm, filters, kernels

Procedia PDF Downloads 180

Authors: Yuan-Hsiang Chang, Pin-Chi Lin, Li-Der Jeng

Abstract:

Advance in techniques of image and video processing has enabled the development of intelligent video surveillance systems. This study was aimed to automatically detect moving human objects and to analyze events of dual human interaction in a surveillance scene. Our system was developed in four major steps: image preprocessing, human object detection, human object tracking, and motion trajectory analysis. The adaptive background subtraction and image processing techniques were used to detect and track moving human objects. To solve the occlusion problem during the interaction, the Kalman filter was used to retain a complete trajectory for each human object. Finally, the motion trajectory analysis was developed to distinguish between the interaction and non-interaction events based on derivatives of trajectories related to the speed of the moving objects. Using a database of 60 video sequences, our system could achieve the classification accuracy of 80% in interaction events and 95% in non-interaction events, respectively. In summary, we have explored the idea to investigate a system for the automatic classification of events for interaction and non-interaction events using surveillance cameras. Ultimately, this system could be incorporated in an intelligent surveillance system for the detection and/or classification of abnormal or criminal events (e.g., theft, snatch, fighting, etc.).

Keywords: motion detection, motion tracking, trajectory analysis, video surveillance

Procedia PDF Downloads 541

Authors: Zhongjie Yu, Hancheng Yu

Abstract:

In this paper, we propose an effective method to detect concentric circles with imperfect edges. First, the gradient of edge pixel is coded and a 2-D lookup table is built to speed up normal generation. Then we take an accumulator to estimate the rough center and collect plausible edges of concentric circles through gradient and distance. Later, we take the contour-based method, which takes the contour and edge intersection, to pre-classify the edges. Finally, we use the extended RANSAC method to find all the candidate circles. The center of concentric circles is determined by the two circles with the highest concentricity. Experimental results demonstrate that the proposed method has both good performance and accuracy for the detection of concentric circles.

Keywords: concentric circle detection, gradient, contour, edge pre-classification, RANSAC

Procedia PDF Downloads 127

Authors: İsmail Kavdır, M. Burak Büyükcan, Ferhat Kurtulmuş

Abstract:

Table olive is a valuable product especially in Mediterranean countries. It is usually consumed after some fermentation process. Defects happened naturally or as a result of an impact while olives are still fresh may become more distinct after processing period. Defected olives are not desired both in table olive and olive oil industries as it will affect the final product quality and reduce market prices considerably. Therefore it is critical to sort table olives before processing or even after processing according to their quality and surface defects. However, doing manual sorting has many drawbacks such as high expenses, subjectivity, tediousness and inconsistency. Quality criterions for green olives were accepted as color and free of mechanical defects, wrinkling, surface blemishes and rotting. In this study, it was aimed to classify fresh table olives using different classifiers and NIR spectroscopy readings and also to compare the classifiers. For this purpose, green (Ayvalik variety) olives were classified based on their surface feature properties such as defect-free, with bruised defect and with fly defect using FT-NIR spectroscopy and classification algorithms such as artificial neural networks, ident and cluster. Bruker multi-purpose analyzer (MPA) FT-NIR spectrometer (Bruker Optik, GmbH, Ettlingen Germany) was used for spectral measurements. The spectrometer was equipped with InGaAs detectors (TE-InGaAs internal for reflectance and RT-InGaAs external for transmittance) and a 20-watt high intensity tungsten–halogen NIR light source. Reflectance measurements were performed with a fiber optic probe (type IN 261) which covered the wavelengths between 780–2500 nm, while transmittance measurements were performed between 800 and 1725 nm. Thirty-two scans were acquired for each reflectance spectrum in about 15.32 s while 128 scans were obtained for transmittance in about 62 s. Resolution was 8 cm⁻¹ for both spectral measurement modes. Instrument control was done using OPUS software (Bruker Optik, GmbH, Ettlingen Germany). Classification applications were performed using three classifiers; Backpropagation Neural Networks, ident and cluster classification algorithms. For these classification applications, Neural Network tool box in Matlab, ident and cluster modules in OPUS software were used. Classifications were performed considering different scenarios; two quality conditions at once (good vs bruised, good vs fly defect) and three quality conditions at once (good, bruised and fly defect). Two spectrometer readings were used in classification applications; reflectance and transmittance. Classification results obtained using artificial neural networks algorithm in discriminating good olives from bruised olives, from olives with fly defect and from the olive group including both bruised and fly defected olives with success rates respectively changing between 97 and 99%, 61 and 94% and between 58.67 and 92%. On the other hand, classification results obtained for discriminating good olives from bruised ones and also for discriminating good olives from fly defected olives using the ident method ranged between 75-97.5% and 32.5-57.5%, respectfully; results obtained for the same classification applications using the cluster method ranged between 52.5-97.5% and between 22.5-57.5%.

Keywords: artificial neural networks, statistical classifiers, NIR spectroscopy, reflectance, transmittance

Procedia PDF Downloads 239

Authors: Magdalena Igras, Joanna Grzybowska, Mariusz Ziółko

Abstract:

The study of emotions expressed in emergency phone call is presented, covering both statistical analysis of emotions configurations and an attempt to automatically classify emotions. An emergency call is a situation usually accompanied by intense, authentic emotions. They influence (and may inhibit) the communication between caller and responder. In order to support responders in their responsible and psychically exhaustive work, we studied when and in which combinations emotions appeared in calls. A corpus of 45 hours of conversations (about 3300 calls) from emergency call center was collected. Each recording was manually tagged with labels of emotions valence (positive, negative or neutral), type (sadness, tiredness, anxiety, surprise, stress, anger, fury, calm, relief, compassion, satisfaction, amusement, joy) and arousal (weak, typical, varying, high) on the basis of perceptual judgment of two annotators. As we concluded, basic emotions tend to appear in specific configurations depending on the overall situational context and attitude of speaker. After performing statistical analysis we distinguished four main types of emotional behavior of callers: worry/helplessness (sadness, tiredness, compassion), alarm (anxiety, intense stress), mistake or neutral request for information (calm, surprise, sometimes with amusement) and pretension/insisting (anger, fury). The frequency of profiles was respectively: 51%, 21%, 18% and 8% of recordings. A model of presenting the complex emotional profiles on the two-dimensional (tension-insecurity) plane was introduced. In the stage of acoustic analysis, a set of prosodic parameters, as well as Mel-Frequency Cepstral Coefficients (MFCC) were used. Using these parameters, complex emotional states were modeled with machine learning techniques including Gaussian mixture models, decision trees and discriminant analysis. Results of classification with several methods will be presented and compared with the state of the art results obtained for classification of basic emotions. Future work will include optimization of the algorithm to perform in real time in order to track changes of emotions during a conversation.

Keywords: acoustic analysis, complex emotions, emotion recognition, machine learning

Procedia PDF Downloads 390

Authors: S. Kherchaoui, A. Houacine

Abstract:

This paper presents a facial expression recognition system. It performs identification and classification of the seven basic expressions; happy, surprise, fear, disgust, sadness, anger, and neutral states. It consists of three main parts. The first one is the detection of a face and the corresponding facial features to extract the most expressive portion of the face, followed by a normalization of the region of interest. Then calculus of curvelet coefficients is performed with dimensionality reduction through principal component analysis. The resulting coefficients are combined with two ratios; mouth ratio and face edge ratio to constitute the whole feature vector. The third step is the classification of the emotional state using the SVM method in the feature space.

Keywords: facial expression identification, curvelet coefficient, support vector machine (SVM), recognition system

Procedia PDF Downloads 229

Authors: Nassima Noufail, Sara Bouhali

Abstract:

In this work, we develop a semi-supervised solution for the purpose of action detection in videos and propose an efficient algorithm for video segmentation. The approach is divided into video segmentation, feature extraction, and classification. In the first part, a video is segmented into clips, and we used the K-means algorithm for this segmentation; our goal is to find groups based on similarity in the video. The application of k-means clustering into all the frames is time-consuming; therefore, we started by the identification of transition frames where the scene in the video changes significantly, and then we applied K-means clustering into these transition frames. We used two image filters, the gaussian filter and the Laplacian of Gaussian. Each filter extracts a set of features from the frames. The Gaussian filter blurs the image and omits the higher frequencies, and the Laplacian of gaussian detects regions of rapid intensity changes; we then used this vector of filter responses as an input to our k-means algorithm. The output is a set of cluster centers. Each video frame pixel is then mapped to the nearest cluster center and painted with a corresponding color to form a visual map. The resulting visual map had similar pixels grouped. We then computed a cluster score indicating how clusters are near each other and plotted a signal representing frame number vs. clustering score. Our hypothesis was that the evolution of the signal would not change if semantically related events were happening in the scene. We marked the breakpoints at which the root mean square level of the signal changes significantly, and each breakpoint is an indication of the beginning of a new video segment. In the second part, for each segment from part 1, we randomly selected a 16-frame clip, then we extracted spatiotemporal features using convolutional 3D network C3D for every 16 frames using a pre-trained model. The C3D final output is a 512-feature vector dimension; hence we used principal component analysis (PCA) for dimensionality reduction. The final part is the classification. The C3D feature vectors are used as input to a multi-class linear support vector machine (SVM) for the training model, and we used a multi-classifier to detect the action. We evaluated our experiment on the UCF101 dataset, which consists of 101 human action categories, and we achieved an accuracy that outperforms the state of art by 1.2%.

Keywords: video segmentation, action detection, classification, Kmeans, C3D

Procedia PDF Downloads 70

Authors: K. Pankaj, R. K. Chaudhary, B. P. Mishra, S. Kochar

Abstract:

Introduction: Consultation-Liaison psychiatry is a branch of psychiatry that includes clinical service, teaching and research. A consultation-liaison psychiatrist plays a role in having an expert opinion and linking the patients to other medical professionals and the patient’s bio-psycho-social aspects that may be leading to his/her symptoms. Consultation-Liaison psychiatry has been recognised as 'The guardian of the holistic approach to the patient', underlining its pre-eminent role in the management of patients who are admitted in a tertiary care hospital. Aims/ Objectives: The aim of the study was to analyse the utilization of psychiatric services and reasons for referrals in a tertiary care hospital. Materials and Methods: The study was done in a tertiary care hospital. The study included all the cases referred from different Inpatient wards to the psychiatry department for consultation. The study was conducted on 300 patients over a 3 month period. International classification of diseases 10 was used to diagnose the referred cases. Results: The majority of the referral was from the Medical Intensive care unit (22%) followed by general medical wards (18.66%). Majority of the referral was taken for altered sensorium (24.66%), followed by low mood or unexplained medical symptoms (21%). Majority of the referrals had a diagnosis of alcohol withdrawal syndrome (21%) as per International classification of diseases criteria, followed by unipolar Depression and Anxiety disorder (~ 14%), followed by Schizophrenia (5%) and Polysubstance abuse (2.6%). Conclusions: Our study concludes the importance of utilization of consultation-liaison psychiatric services. Also, the study signifies the need for sensitization of our colleagues regarding psychiatric sign and symptoms from time to time and seek psychiatric consult timely to decrease morbidity.

Keywords: consultation-liaison, psychiatry, referral, tertiary care hospital

Procedia PDF Downloads 146