Our research focuses on the crucial challenge of discerning text produced by Large Language Models (LLMs) from human-generated text, which holds significance for various applications. With ongoing discussions about attaining a model with such functionality, we present supporting evidence regarding the feasibility of such models. We evaluated our models on multiple datasets, including Twitter Sentiment, Football Commentary, Project Gutenberg, PubMedQA, and SQuAD, confirming the efficacy of the enhanced detection approaches. These datasets were sampled with intricate constraints encompassing every possibility, laying the foundation for future research. We evaluate GPT-3.5-Turbo against various detectors such as SVM, RoBERTa-base, and RoBERTa-large. Based on the research findings, the results predominantly relied on the sequence length of the sentence.
Although statistical learning theory provides a robust framework to understand supervised learning, many theoretical aspects of deep learning remain unclear, in particular how different architectures may lead to inductive bias when trained using gradient based methods. The goal of these lectures is to provide an overview of some of the main questions that arise when attempting to understand deep learning from a learning theory perspective. After a brief reminder on statistical learning theory and stochastic optimization, we discuss implicit bias in the context of benign overfitting. We then move to a general description of the mirror descent algorithm, showing how we may go back and forth between a parameter space and the corresponding function space for a given learning problem, as well as how the geometry of the learning problem may be represented by a metric tensor. Building on this framework, we provide a detailed study of the implicit bias of gradient descent on linear diagonal networks for various regression tasks, showing how the loss function, scale of parameters at initialization and depth of the network may lead to various forms of implicit bias, in particular transitioning between kernel or feature learning.
The collection of data to train a Deep Learning network is costly in terms of effort and resources. In many cases, especially in a medical context, it may have detrimental impacts. Such as requiring invasive medical procedures or processes which could in themselves cause medical harm. However, Deep Learning is seen as a data hungry method. Here, we look at two commonly held adages i) more data gives better results and ii) transfer learning will aid you when you don't have enough data. These are widely assumed to be true and used as evidence for choosing how to solve a problem when Deep Learning is involved. We evaluate six medical datasets and six general datasets. Training a ResNet18 network on varying subsets of these datasets to evaluate `more data gives better results'. We take eleven of these datasets as the sources for Transfer Learning on subsets of the twelfth dataset -- Chest -- in order to determine whether Transfer Learning is universally beneficial. We go further to see whether multi-stage Transfer Learning provides a consistent benefit. Our analysis shows that the real situation is more complex than these simple adages -- more data could lead to a case of diminishing returns and an incorrect choice of dataset for transfer learning can lead to worse performance, with datasets which we would consider highly similar to the Chest dataset giving worse results than datasets which are more dissimilar. Multi-stage transfer learning likewise reveals complex relationships between datasets.
Fingerprints are widely recognized as one of the most unique and reliable characteristics of human identity. Most modern fingerprint authentication systems rely on contact-based fingerprints, which require the use of fingerprint scanners or fingerprint sensors for capturing fingerprints during the authentication process. Various types of fingerprint sensors, such as optical, capacitive, and ultrasonic sensors, employ distinct techniques to gather and analyze fingerprint data. This dependency on specific hardware or sensors creates a barrier or challenge for the broader adoption of fingerprint based biometric systems. This limitation hinders the widespread adoption of fingerprint authentication in various applications and scenarios. Border control, healthcare systems, educational institutions, financial transactions, and airport security face challenges when fingerprint sensors are not universally available. To mitigate the dependence on additional hardware, the use of contactless fingerprints has emerged as an alternative. Developing precise fingerprint segmentation methods, accurate fingerprint extraction tools, and reliable fingerprint matchers are crucial for the successful implementation of a robust contactless fingerprint authentication system. This paper focuses on the development of a deep learning-based segmentation tool for contactless fingerprint localization and segmentation. Our system leverages deep learning techniques to achieve high segmentation accuracy and reliable extraction of fingerprints from contactless fingerprint images. In our evaluation, our segmentation method demonstrated an average mean absolute error (MAE) of 30 pixels, an error in angle prediction (EAP) of 5.92 degrees, and a labeling accuracy of 97.46%. These results demonstrate the effectiveness of our novel contactless fingerprint segmentation and extraction tools.
Uncertainty quantification is vital for safety-critical Deep Learning applications like medical image segmentation. We introduce BA U-Net, an uncertainty-aware model for MRI segmentation that integrates Bayesian Neural Networks with Attention Mechanisms. BA U-Net delivers accurate, interpretable results, crucial for reliable pathology screening. Evaluated on BraTS 2020, this model addresses the critical need for confidence estimation in deep learning-based medical imaging.
Atherosclerosis, a chronic inflammatory disease affecting the large arteries, presents a global health risk. Accurate analysis of diagnostic images, like computed tomographic angiograms (CTAs), is essential for staging and monitoring the progression of atherosclerosis-related conditions, including peripheral arterial disease (PAD). However, manual analysis of CTA images is time-consuming and tedious. To address this limitation, we employed a deep learning model to segment the vascular system in CTA images of PAD patients undergoing femoral endarterectomy surgery and to measure vascular calcification from the left renal artery to the patella. Utilizing proprietary CTA images of 27 patients undergoing femoral endarterectomy surgery provided by Prisma Health Midlands, we developed a Deep Neural Network (DNN) model to first segment the arterial system, starting from the descending aorta to the patella, and second, to provide a metric of arterial calcification. Our designed DNN achieved 83.4% average Dice accuracy in segmenting arteries from aorta to patella, advancing the state-of-the-art by 0.8%. Furthermore, our work is the first to present a robust statistical analysis of automated calcification measurement in the lower extremities using deep learning, attaining a Mean Absolute Percentage Error (MAPE) of 9.5% and a correlation coefficient of 0.978 between automated and manual calcification scores. These findings underscore the potential of deep learning techniques as a rapid and accurate tool for medical professionals to assess calcification in the abdominal aorta and its branches above the patella. The developed DNN model and related documentation in this project are available at GitHub page at https://github.com/pip-alireza/DeepCalcScoring.
Marine snow, the floating particles in underwater images, severely degrades the visibility and performance of human and machine vision systems. This paper proposes a novel method to reduce the marine snow interference using deep learning techniques. We first synthesize realistic marine snow samples by training a Generative Adversarial Network (GAN) model and combine them with natural underwater images to create a paired dataset. We then train a U-Net model to perform marine snow removal as an image to image translation task. Our experiments show that the U-Net model can effectively remove both synthetic and natural marine snow with high accuracy, outperforming state-of-the-art methods such as the Median filter and its adaptive variant. We also demonstrate the robustness of our method by testing it on the MSRB dataset, which contains synthetic artifacts that our model has not seen during training. Our method is a practical and efficient solution for enhancing underwater images affected by marine snow.
Deep learning has revolutionized the design of the channel state information (CSI) feedback module in wireless communications. However, designing the optimal neural network (NN) architecture for CSI feedback can be a laborious and time-consuming process. Manual design can be prohibitively expensive for customizing NNs to different scenarios. This paper proposes using neural architecture search (NAS) to automate the generation of scenario-customized CSI feedback NN architectures, thereby maximizing the potential of deep learning in exclusive environments. By employing automated machine learning and gradient-descent-based NAS, an efficient and cost-effective architecture design process is achieved. The proposed approach leverages implicit scene knowledge, integrating it into the scenario customization process in a data-driven manner, and fully exploits the potential of deep learning for each specific scenario. To address the issue of excessive search, early stopping and elastic selection mechanisms are employed, enhancing the efficiency of the proposed scheme. The experimental results demonstrate that the automatically generated architecture, known as Auto-CsiNet, outperforms manually-designed models in both reconstruction performance (achieving approximately a 14% improvement) and complexity (reducing it by approximately 50%). Furthermore, the paper analyzes the impact of the scenario on the NN architecture and its capacity.
Artificial intelligence (AI) has emerged as a transformative influence, engendering paradigm shifts in global societies, spanning academia and industry. However, in light of these rapid advances, addressing the underrepresentation of black communities and African countries in AI is crucial. Boosting enthusiasm for AI can be effectively accomplished by showcasing straightforward applications around tasks like identifying and categorizing traditional symbols, such as Adinkra symbols, or familiar objects within the community. In this research endeavor, we dived into classical machine learning and harnessed the power of deep learning models to tackle the intricate task of classifying and recognizing Adinkra symbols. The idea led to a newly constructed ADINKRA dataset comprising 174,338 images meticulously organized into 62 distinct classes, each representing a singular and emblematic symbol. We constructed a CNN model for classification and recognition using six convolutional layers, three fully connected (FC) layers, and optional dropout regularization. The model is a simpler and smaller version of VGG, with fewer layers, smaller channel sizes, and a fixed kernel size. Additionally, we tap into the transfer learning capabilities provided by pre-trained models like VGG and ResNet. These models assist us in both classifying images and extracting features that can be used with classical machine learning models. We assess the model's performance by measuring its accuracy and convergence rate and visualizing the areas that significantly influence its predictions. These evaluations serve as a foundational benchmark for future assessments of the ADINKRA dataset. We hope this application exemplar inspires ideas on the various uses of AI in organizing our traditional and modern lives.
In the context of deep learning research, where model introductions continually occur, the need for effective and efficient evaluation remains paramount. Existing methods often emphasize accuracy metrics, overlooking stability. To address this, the paper introduces the Accuracy-Stability Index (ASI), a quantitative measure incorporating both accuracy and stability for assessing deep learning models. Experimental results demonstrate the application of ASI, and a 3D surface model is presented for visualizing ASI, mean accuracy, and coefficient of variation. This paper addresses the important issue of quantitative benchmarking metrics for deep learning models, providing a new approach for accurately evaluating accuracy and stability of deep learning models. The paper concludes with discussions on potential weaknesses and outlines future research directions.
The concept of deep learning is employed for the inversion of NMR signals and it is shown that NMR signal inversion can be considered as an image-to-image regression problem, which can be treated with a convolutional neural net. It is further outlined, that inversion through deep learning provides a clear efficiency and usability advantage compared to regularization techniques such as Tikhonov and modified total generalized variation (MTGV), because no hyperparemeter selection prior to reconstruction is necessary. The inversion network is applied to simulated NMR signals and the results compared with Tikhonov- and MTGV-regularization. The comparison shows that inversion via deep learning is significantly faster than the latter regularization methods and also outperforms both regularization techniques in nearly all instances.
Label-free approaches are attractive in cytological imaging due to their flexibility and cost efficiency. They are supported by machine learning methods, which, despite the lack of labeling and the associated lower contrast, can classify cells with high accuracy where the human observer has little chance to discriminate cells. In order to better integrate these workflows into the clinical decision making process, this work investigates the calibration of confidence estimation for the automated classification of leukocytes. In addition, different visual explanation approaches are compared, which should bring machine decision making closer to professional healthcare applications. Furthermore, we were able to identify general detection patterns in neural networks and demonstrate the utility of the presented approaches in different scenarios of blood cell analysis.
In recent years, a variety of deep learning networks for cardiac MRI (CMR) segmentation have been developed and analyzed. However, nearly all of them are focused on cine CMR under breathold. In this work, accuracy of deep learning methods is assessed for volumetric analysis (via segmentation) of the left ventricle in real-time free-breathing CMR at rest and under exercise stress. Data from healthy volunteers (n=15) for cine and real-time free-breathing CMR at rest and under exercise stress were analyzed retrospectively. Segmentations of a commercial software (comDL) and a freely available neural network (nnU-Net), were compared to a reference created via the manual correction of comDL segmentation. Segmentation of left ventricular endocardium (LV), left ventricular myocardium (MYO), and right ventricle (RV) is evaluated for both end-systolic and end-diastolic phases and analyzed with Dice's coefficient (DC). The volumetric analysis includes LV end-diastolic volume (EDV), LV end-systolic volume (ESV), and LV ejection fraction (EF). For cine CMR, nnU-Net and comDL achieve a DC above 0.95 for LV and 0.9 for MYO, and RV. For real-time CMR, the accuracy of nnU-Net exceeds that of comDL overall. For real-time CMR at rest, nnU-Net achieves a DC of 0.94 for LV, 0.89 for MYO, and 0.90 for RV; mean absolute differences between nnU-Net and reference are 2.9mL for EDV, 3.5mL for ESV and 2.6% for EF. For real-time CMR under exercise stress, nnU-Net achieves a DC of 0.92 for LV, 0.85 for MYO, and 0.83 for RV; mean absolute differences between nnU-Net and reference are 11.4mL for EDV, 2.9mL for ESV and 3.6% for EF. Deep learning methods designed or trained for cine CMR segmentation can perform well on real-time CMR. For real-time free-breathing CMR at rest, the performance of deep learning methods is comparable to inter-observer variability in cine CMR and is usable or fully automatic segmentation.
It is crucial for emergency physicians to identify patients at higher risk of mortality to effectively prioritize hospital resources, particularly in regions with limited medical services. This became even more critical during global pandemics, which have disrupted lives in unprecedented ways and caused widespread morbidity and mortality. The collected data from patients is beneficial to predict the outcome, although there is a question about which data makes the most accurate predictions. Therefore, this study aimed to achieve two main objectives during the pandemic, using data and experiments from the most recent global health crisis, COVID-19. First, we want to examine whether deep learning algorithms can predict a patient's morality. Second, we investigated the impact of Clinical and RT-PCR on prediction to determine which one is more reliable. We defined four stages with different feature sets and used 9 machine learning and deep learning methods to build appropriate model. Based on results, the deep neural decision forest, as an interpretable deep learning methods, performed the best across all stages and proved its capability to predict the recovery and death of patients. Additionally, results indicate that Clinical alone (without the use of RT-PCR) is the most effective method of diagnosis, with an accuracy of 80%. This study can provide guidance for medical professionals in the event of a crisis or outbreak similar to COVID-19. Moreover, the proposed deep learning method demonstrates exceptional suitability for mortality prediction.
These days, deep learning models have achieved great success in multiple fields, from autonomous driving to medical diagnosis. These models have expanded the abilities of artificial intelligence by offering great solutions to complex problems that were very difficult to solve earlier. In spite of their unseen success in various, it has been identified, through research conducted, that deep learning models can be subjected to various attacks that compromise model security and data privacy of the Deep Neural Network models. Deep learning models can be subjected to various attacks at different stages of their lifecycle. During the testing phase, attackers can exploit vulnerabilities through different kinds of attacks such as Model Extraction Attacks, Model Inversion attacks, and Adversarial attacks. Model Extraction Attacks are aimed at reverse-engineering a trained deep learning model, with the primary objective of revealing its architecture and parameters. Model inversion attacks aim to compromise the privacy of the data used in the Deep learning model. These attacks are done to compromise the confidentiality of the model by going through the sensitive training data from the model's predictions. By analyzing the model's responses, attackers aim to reconstruct sensitive information. In this way, the model's data privacy is compromised. Adversarial attacks, mainly employed on computer vision models, are made to corrupt models into confidently making incorrect predictions through malicious testing data. These attacks subtly alter the input data, making it look normal but misleading deep learning models to make incorrect decisions. Such attacks can happen during both the model's evaluation and training phases. Data Poisoning Attacks add harmful data to the training set, disrupting the learning process and reducing the reliability of the deep learning mode.
Data sampling is an effective method to improve the training speed of neural networks, with recent results demonstrating that it can even break the neural scaling laws. These results critically rely on high-quality scores to estimate the importance of an input to the network. We observe that there are two dominant strategies: static sampling, where the scores are determined before training, and dynamic sampling, where the scores can depend on the model weights. Static algorithms are computationally inexpensive but less effective than their dynamic counterparts, which can cause end-to-end slowdown due to their need to explicitly compute losses. To address this problem, we propose a novel sampling distribution based on nonparametric kernel regression that learns an effective importance score as the neural network trains. However, nonparametric regression models are too computationally expensive to accelerate end-to-end training. Therefore, we develop an efficient sketch-based approximation to the Nadaraya-Watson estimator. Using recent techniques from high-dimensional statistics and randomized algorithms, we prove that our Nadaraya-Watson sketch approximates the estimator with exponential convergence guarantees. Our sampling algorithm outperforms the baseline in terms of wall-clock time and accuracy on four datasets.
Splice sites play a crucial role in gene expression, and accurate prediction of these sites in DNA sequences is essential for diagnosing and treating genetic disorders. We address the challenge of splice site prediction by introducing DeepDeCode, an attention-based deep learning sequence model to capture the long-term dependencies in the nucleotides in DNA sequences. We further propose using visualization techniques for accurate identification of sequence motifs, which enhance the interpretability and trustworthiness of DeepDeCode. We compare DeepDeCode to other state-of-the-art methods for splice site prediction and demonstrate its accuracy, explainability and efficiency. Given the results of our methodology, we expect that it can used for healthcare applications to reason about genomic processes and be extended to discover new splice sites and genomic regulatory elements.
While deep learning has achieved remarkable success, there is no clear explanation about why it works so well. In order to discuss this question quantitatively, we need a mathematical framework that explains what learning is in the first place. After several considerations, we succeeded in constructing a mathematical framework that can provide a unified understanding of all types of learning, including deep learning and learning in the brain. We call it learning principle, and it follows that all learning is equivalent to estimating the probability of input data. We not only derived this principle, but also mentioned its application to actual machine learning models. For example, we found that conventional supervised learning is equivalent to estimating conditional probabilities, and succeeded in making supervised learning more effective and generalized. We also proposed a new method of defining the values of estimated probability using differentiation, and showed that unsupervised learning can be performed on arbitrary dataset without any prior knowledge. Namely, this method is a general-purpose machine learning in the true sense. Moreover, we succeeded in describing the learning mechanism in the brain by considering the time evolution of a fully or partially connected model and applying this new method. The learning principle provides solutions to many unsolved problems in deep learning and cognitive neuroscience.
This paper introduces GeoMorph, a novel geometric deep-learning framework designed for image registration of cortical surfaces. The registration process consists of two main steps. First, independent feature extraction is performed on each input surface using graph convolutions, generating low-dimensional feature representations that capture important cortical surface characteristics. Subsequently, features are registered in a deep-discrete manner to optimize the overlap of common structures across surfaces by learning displacements of a set of control points. To ensure smooth and biologically plausible deformations, we implement regularization through a deep conditional random field implemented with a recurrent neural network. Experimental results demonstrate that GeoMorph surpasses existing deep-learning methods by achieving improved alignment with smoother deformations. Furthermore, GeoMorph exhibits competitive performance compared to classical frameworks. Such versatility and robustness suggest strong potential for various neuroscience applications.
Bayesian deep learning and conformal prediction are two methods that have been used to convey uncertainty and increase safety in machine learning systems. We focus on combining Bayesian deep learning with split conformal prediction and how this combination effects out-of-distribution coverage; particularly in the case of multiclass image classification. We suggest that if the model is generally underconfident on the calibration set, then the resultant conformal sets may exhibit worse out-of-distribution coverage compared to simple predictive credible sets. Conversely, if the model is overconfident on the calibration set, the use of conformal prediction may improve out-of-distribution coverage. We evaluate prediction sets as a result of combining split conformal methods and neural networks trained with (i) stochastic gradient descent, (ii) deep ensembles, and (iii) mean-field variational inference. Our results suggest that combining Bayesian deep learning models with split conformal prediction can, in some cases, cause unintended consequences such as reducing out-of-distribution coverage.