Browsing by Author "Badirli, Sarkhan"

Now showing 1 - 9 of 9
  • Thumbnail Image
    Item
    Bayesian Zero-Shot Learning
    (Springer, 2020) Badirli, Sarkhan; Akata, Zeynep; Dundar, Murat; Computer and Information Science, School of Science
    Object classes that surround us have a natural tendency to emerge at varying levels of abstraction. We propose a Bayesian approach to zero-shot learning (ZSL) that introduces the notion of meta-classes and implements a Bayesian hierarchy around these classes to effectively blend data likelihood with local and global priors. Local priors driven by data from seen classes, i.e., classes available at training time, become instrumental in recovering unseen classes, i.e., classes that are missing at training time, in a generalized ZSL (GZSL) setting. Hyperparameters of the Bayesian model offer a convenient way to optimize the trade-off between seen and unseen class accuracy. We conduct experiments on seven benchmark datasets, including a large scale ImageNet and show that our model produces promising results in the challenging GZSL setting.
  • Thumbnail Image
    Item
    Classifying the Unknown: Identification of Insects by Deep Open-set Bayesian Learning
    (bioRxiv, 2021-09-17) Badirli, Sarkhan; Picard, Christine J.; Mohler, George; Akata, Zeynep; Dundar, Murat
    Insects represent a large majority of biodiversity on Earth, yet only 20% of the estimated 5.5 million insect species are currently described (1). While describing new species typically requires specific taxonomic expertise to identify morphological characters that distinguish it from other potential species, DNA-based methods have aided in providing additional evidence of separate species (2). Machine learning (ML) is emerging as a potential new approach in identifying new species, given that this analysis may be more sensitive to subtle differences humans may not process. Existing ML algorithms are limited by image repositories that do not include undescribed species. We developed a Bayesian deep learning method for the open-set classification of species. The proposed approach forms a Bayesian hierarchy of species around corresponding genera and uses deep embeddings of images and barcodes together to identify insects at the lowest level of abstraction possible. To demonstrate proof of concept, we used a database of 32,848 insect instances from 1,040 described species split into training and test data. The test data included 243 species not present in the training data. Our results demonstrate that using DNA sequences and images together, insect instances of described species can be classified with 96.66% accuracy while achieving accuracy of 81.39% in identifying genera of insect instances of undescribed species. The proposed deep open-set Bayesian model demonstrates a powerful new approach that can be used for the gargantuan task of identifying new insect species.
  • Thumbnail Image
    Item
    Classifying the unknown: Insect identification with deep hierarchical Bayesian learning
    (Wiley, 2023) Badirli, Sarkhan; Picard, Christine Johanna; Mohler, George; Richert, Frannie; Akata, Zeynep; Dundar, Murat
    1. Classifying insect species involves a tedious process of identifying distinctive morphological insect characters by taxonomic experts. Machine learning can harness the power of computers to potentially create an accurate and efficient method for performing this task at scale, given that its analytical processing can be more sensitive to subtle physical differences in insects, which experts may not perceive. However, existing machine learning methods are designed to only classify insect samples into described species, thus failing to identify samples from undescribed species. 2. We propose a novel deep hierarchical Bayesian model for insect classification, given the taxonomic hierarchy inherent in insects. This model can classify samples of both described and undescribed species; described samples are assigned a species while undescribed samples are assigned a genus, which is a pivotal advancement over just identifying them as outliers. We demonstrated this proof of concept on a new database containing paired insect image and DNA barcode data from four insect orders, including 1040 species, which far exceeds the number of species used in existing work. A quarter of the species were excluded from the training set to simulate undescribed species. 3. With the proposed classification framework using combined image and DNA data in the model, species classification accuracy for described species was 96.66% and genus classification accuracy for undescribed species was 81.39%. Including both data sources in the model resulted in significant improvement over including image data only (39.11% accuracy for described species and 35.88% genus accuracy for undescribed species), and modest improvement over including DNA data only (73.39% genus accuracy for undescribed species). 4. Unlike current machine learning methods, the proposed deep hierarchical Bayesian learning approach can simultaneously classify samples of both described and undescribed species, a functionality that could become instrumental in biodiversity monitoring across the globe. This framework can be customized for any taxonomic classification problem for which image and DNA data can be obtained, thus making it relevant for use across all biological kingdoms.
  • Thumbnail Image
    Item
    Coupled IGMM-GANs for improved generative adversarial anomaly detection
    (IEEE, 2018-12) Gray, Kathryn; Smolyak, Daniel; Badirli, Sarkhan; Mohler, George; Computer and Information Science, School of Science
    Detecting anomalies and outliers in data has a number of applications including hazard sensing, fraud detection, and systems management. While generative adversarial networks seem like a natural fit for addressing these challenges, we find that existing GAN based anomaly detection algorithms perform poorly due to their inability to handle multimodal patterns. For this purpose we introduce an infinite Gaussian mixture model coupled with (bi-directional) generative adversarial networks, IGMM-GAN, that facilitates multimodal anomaly detection. We illustrate our methodology and its improvement over existing GAN anomaly detection on the MNIST dataset.
  • Thumbnail Image
    Item
    Coupled IGMM-GANs with Applications to Anomaly Detection in Human Mobility Data
    (ACM, 2020-12) Smolyak, Daniel; Gray, Kathryn; Badirli, Sarkhan; Mohler, George; Computer and Information Science, School of Science
    Detecting anomalous activity in human mobility data has a number of applications, including road hazard sensing, telematics-based insurance, and fraud detection in taxi services and ride sharing. In this article, we address two challenges that arise in the study of anomalous human trajectories: (1) a lack of ground truth data on what defines an anomaly and (2) the dependence of existing methods on significant pre-processing and feature engineering. Although generative adversarial networks (GANs) seem like a natural fit for addressing these challenges, we find that existing GAN-based anomaly detection algorithms perform poorly due to their inability to handle multimodal patterns. For this purpose, we introduce an infinite Gaussian mixture model coupled with (bidirectional) GANs—IGMM-GAN—that is able to generate synthetic, yet realistic, human mobility data and simultaneously facilitates multimodal anomaly detection. Through the estimation of a generative probability density on the space of human trajectories, we are able to generate realistic synthetic datasets that can be used to benchmark existing anomaly detection methods. The estimated multimodal density also allows for a natural definition of outlier that we use for detecting anomalous trajectories. We illustrate our methodology and its improvement over existing GAN anomaly detection on several human mobility datasets, along with MNIST.
  • Thumbnail Image
    Item
    Fine-Grained Zero-Shot Learning with DNA as Side Information
    (NeurIPS 2021, 2021-09-29) Badirli, Sarkhan; Akata, Zeynep; Mohler, George; Picard, Christine J.; Dundar, Murat; Biology, School of Science
    Fine-grained zero-shot learning task requires some form of side-information to transfer discriminative information from seen to unseen classes. As manually annotated visual attributes are extremely costly and often impractical to obtain for a large number of classes, in this study we use DNA as side information for the first time for fine-grained zero-shot classification of species. Mitochondrial DNA plays an important role as a genetic marker in evolutionary biology and has been used to achieve near-perfect accuracy in the species classification of living organisms. We implement a simple hierarchical Bayesian model that uses DNA information to establish the hierarchy in the image space and employs local priors to define surrogate classes for unseen ones. On the benchmark CUB dataset, we show that DNA can be equally promising yet in general a more accessible alternative than word vectors as a side information. This is especially important as obtaining robust word representations for fine-grained species names is not a practicable goal when information about these species in free-form text is limited. On a newly compiled fine-grained insect dataset that uses DNA information from over a thousand species, we show that the Bayesian approach outperforms state-of-the-art by a wide margin.
  • Thumbnail Image
    Item
    Open Set Authorship Attribution Toward Demystifying Victorian Periodicals
    (Springer, 2021-09) Badirli, Sarkhan; Borgo Ton, Mary; Gungor, Abdulmecit; Dundar, Murat; Computer and Information Science, School of Science
    Existing research in computational authorship attribution (AA) has primarily focused on attribution tasks with a limited number of authors in a closed-set configuration. This restricted set-up is far from being realistic in dealing with highly entangled real-world AA tasks that involve a large number of candidate authors for attribution during test time. In this paper, we study AA in historical texts using a new data set compiled from the Victorian literature. We investigate the predictive capacity of most common English words in distinguishing writings of most prominent Victorian novelists. We challenged the closed-set classification assumption and discussed the limitations of standard machine learning techniques in dealing with the open set AA task. Our experiments suggest that a linear classifier can achieve near perfect attribution accuracy under closed set assumption yet, the need for more robust approaches becomes evident once a large candidate pool has to be considered in the open-set classification setting.
  • Thumbnail Image
    Item
    Prediction of Pubertal Mandibular Growth in Males with Class II Malocclusion by Utilizing Machine Learning
    (MDPI, 2023-08-21) Zakhar, Grant; Hazime, Samir; Eckert, George; Wong, Ariel; Badirli, Sarkhan; Turkkahraman, Hakan; Orthodontics and Oral Facial Genetics, School of Dentistry
    The goal of this study was to create a novel machine learning (ML) model that can predict the magnitude and direction of pubertal mandibular growth in males with Class II malocclusion. Lateral cephalometric radiographs of 123 males at three time points (T1: 12; T2: 14; T3: 16 years old) were collected from an online database of longitudinal growth studies. Each radiograph was traced, and seven different ML models were trained using 38 data points obtained from 92 subjects. Thirty-one subjects were used as the test group to predict the post-pubertal mandibular length and y-axis, using input data from T1 and T2 combined (2 year prediction), and T1 alone (4 year prediction). Mean absolute errors (MAEs) were used to evaluate the accuracy of each model. For all ML methods tested using the 2 year prediction, the MAEs for post-pubertal mandibular length ranged from 2.11–6.07 mm to 0.85–2.74° for the y-axis. For all ML methods tested with 4 year prediction, the MAEs for post-pubertal mandibular length ranged from 2.32–5.28 mm to 1.25–1.72° for the y-axis. Besides its initial length, the most predictive factors for mandibular length were found to be chronological age, upper and lower face heights, upper and lower incisor positions, and inclinations. For the y-axis, the most predictive factors were found to be y-axis at earlier time points, SN-MP, SN-Pog, SNB, and SNA. Although the potential of ML techniques to accurately forecast future mandibular growth in Class II cases is promising, a requirement for more substantial sample sizes exists to further enhance the precision of these predictions.
  • Thumbnail Image
    Item
    Short- and Long-Term Prediction of the Post-Pubertal Mandibular Length and Y-Axis in Females Utilizing Machine Learning
    (MDPI, 2023-08-22) Parrish, Matthew; O’Connell, Ella; Eckert, George; Hughes, Jay; Badirli, Sarkhan; Turkkahraman, Hakan; Orthodontics and Oral Facial Genetics, School of Dentistry
    The aim of this study was to create a novel machine learning (ML) algorithm for predicting the post-pubertal mandibular length and Y-axis in females. Cephalometric data from 176 females with Angle Class I occlusion were used to train and test seven ML algorithms. For all ML methods tested, the mean absolute errors (MAEs) for the 2-year prediction ranged from 2.78 to 5.40 mm and 0.88 to 1.48 degrees, respectively. For the 4-year prediction, MAEs of mandibular length and Y-axis ranged from 3.21 to 4.00 mm and 1.19 to 5.12 degrees, respectively. The most predictive factors for post-pubertal mandibular length were mandibular length at previous timepoints, age, sagittal positions of the maxillary and mandibular skeletal bases, mandibular plane angle, and anterior and posterior face heights. The most predictive factors for post-pubertal Y-axis were Y-axis at previous timepoints, mandibular plane angle, and sagittal positions of the maxillary and mandibular skeletal bases. ML methods were identified as capable of predicting mandibular length within 3 mm and Y-axis within 1 degree. Compared to each other, all of the ML algorithms were similarly accurate, with the exception of multilayer perceptron regressor.