Operational Neural Networks (ONNs) have recently been proposed to address the well-known limitations and drawbacks of conventional Convolutional (CNNs) such as network homogeneity with sole linear neuron model. ONNs are heterogeneous networks a generalized However operator search method in is not only computationally demanding, but heterogeneity also limited since same set operators will then be used for all neurons each layer. Moreover, performance directly depends on library used, which...

Discriminative learning based on convolutional neural networks (CNNs) aims to perform image restoration by from training examples of noisy-clean pairs. It has become the go-to methodology for tackling and outperformed traditional non-local class methods. However, top-performing are generally composed many layers hundreds neurons, with trainable parameters in excess several million. We claim that this is due inherently linear nature convolution-based transformation, which inadequate handling...

<i>Objective:</i> Despitethe proliferation of numerous deep learning methods proposed for generic ECG classification and arrhythmia detection, compact systems with the real-time ability high accuracy classifying patient-specific are still few. Particularly, scarcity data poses an ultimate challenge to any classifier. Recently, 1D Convolutional Neural Networks (CNNs) have achieved <i>state-of-the-art</i> performance level accurate ventricular supraventricular ectopic beats. However, several...

Glaucoma leads to permanent vision disability by damaging the optical nerve that transmits visual images brain. The fact glaucoma does not show any symptoms as it progresses and cannot be stopped at later stages, makes critical diagnosed in its early stages. Although various deep learning models have been applied for detecting from digital fundus images, due scarcity of labeled data, their generalization performance was limited along with high computational complexity special hardware...

ECG recordings often suffer from a set of artifacts with varying types, severities, and durations, this makes an accurate diagnosis by machines or medical doctors difficult unreliable. Numerous studies have proposed denoising; however, they naturally fail to restore the actual signal corrupted such due their simple naive noise model. In pilot study, we propose novel approach for blind restoration using cycle-consistent generative adversarial networks (Cycle-GANs) where quality can be...

Although numerous R-peak detectors have been proposed in the literature, their robustness and performance levels may significantly deteriorate low-quality noisy signals acquired from mobile electrocardiogram (ECG) sensors, such as Holter monitors. Recently, this issue has addressed by deep 1-D convolutional neural networks (CNNs) that achieved state-of-the-art monitors; however, they pose a high complexity level requires special parallelized hardware setup for real-time processing. On other...

Preventive maintenance of modern electric rotating machinery (RM) is critical for ensuring reliable operation, preventing unpredicted breakdowns and avoiding costly repairs. Recently many studies investigated machine learning monitoring methods especially based on Deep Learning networks focusing mostly detecting bearing faults; however, none them addressed fault severity classification early diagnosis with high enough accuracy. 1D Convolutional Neural Networks (CNNs) have indeed achieved...

Echocardiogram (echo) is the earliest and primary tool for identifying regional wall motion abnormalities (RWMA) in order to diagnose myocardial infarction (MI) or commonly known as heart attack. This paper proposes a novel approach, Active Polynomials, which can accurately robustly estimate global of Left Ventricular (LV) from any echo robust accurate way. The proposed algorithm quantifies true occurring LV segments so assist cardiologists early signs an acute MI. It further enables medical...

Abstract The recently proposed network model, Operational Neural Networks (ONNs), can generalize the conventional Convolutional (CNNs) that are homogenous only with a linear neuron model. As heterogenous ONNs based on generalized model encapsulate any set of non-linear operators to boost diversity and learn highly complex multi-modal functions or spaces minimal complexity training data. However, default search method find optimal in ONNs, so-called Greedy Iterative Search (GIS) method,...

In end-to-end optimized learned image compression, it is standard practice to use a convolutional variational autoencoder with generalized divisive normalization (GDN) transform images into latent space. Recently, Operational Neural Networks (ONNs) that learn the best non-linearity from set of alternatives, and their "self-organized" variants, Self-ONNs, approximate any via Taylor series have been proposed address limitations layers fixed nonlinear activation. this paper, we propose replace...

Operational Neural Networks (ONNs) have recently been proposed as a special class of artificial neural networks for grid structured data. They enable heterogenous non-linear operations to generalize the widely adopted convolution-based neuron model. This work introduces fast GPU-enabled library training operational networks, FastONN, which is based on novel vectorized formulation neurons. Leveraging automatic reverse-mode differentiation backpropagation, FastONN enables increased flexibility...

Computer-aided diagnosis (CAD) based on histopathological imaging has progressed rapidly in recent years with the rise of machine learning methodologies. Traditional approaches consist training a classification model using features extracted from images, textures or morphological properties. Recently, deep-learning methods have been applied directly to raw (unprocessed) data. However, their usability is impacted by paucity annotated data biomedical sector. In order leverage capabilities deep...

Although numerous R-peak detectors have been proposed in the literature, their robustness and performance levels may significantly deteriorate low-quality noisy signals acquired from mobile electrocardiogram (ECG) sensors, such as Holter monitors. Recently, this issue has addressed by deep 1-D convolutional neural networks (CNNs) that achieved state-of-the-art monitors; however, they pose a high complexity level requires special parallelized hardware setup for real-time processing. On other...

Self-Organized Operational Neural Networks (Self-ONNs) have recently been proposed as new-generation neural network models with nonlinear learning units, i.e., the generative neurons that yield an elegant level of diversity; however, like its predecessor, conventional Convolutional (CNNs), they still a common drawback: <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">localized</i> (fixed) kernel operations. This severely limits receptive field...

It has become a standard practice to use the convolutional networks (ConvNet) with RELU non-linearity in image restoration and super-resolution (SR). Although universal approximation theorem states that multi-layer neural network can approximate any non-linear function desired precision, it does not reveal best architecture do so. Recently, operational (ONNs) choose from set of alternatives, their "self-organized" variants (Self-ONN) via Taylor series have been proposed address well-known...

Automatic classification of speech commands has revolutionized human computer interactions in robotic applications. However, employed recognition models usually follow the methodology deep learning with complicated networks which are memory and energy hungry. So, there is a need to either squeeze these or use more efficient lightweight order be able implement resulting classifiers on embedded devices. In this paper, we pick second approach propose network layer enhance command capability...

Binary segmentation of volumetric images porous media is a crucial step towards gaining deeper understanding the factors governing biogeochemical processes at minute scales. Contemporary work primarily revolves around primitive techniques based on global or local adaptive thresholding that have known common drawbacks in image segmentation. Moreover, absence unified benchmark prohibits quantitative evaluation, which further undermines impact existing methodologies. In this study, we tackle...

Despite their recent success on image denoising, the need for deep and complex architectures still hinders practical usage of CNNs. Older but computationally more efficient methods such as BM3D remain a popular choice, especially in resource-constrained scenarios. In this study, we aim to find out whether compact neural networks can learn produce competitive results compared AWGN denoising. To end, conFigure with only two hidden layers employ different neuron models layer widths comparing...

Convolutional Neural Networks (CNNs) have recently become a favored technique for image denoising due to its adaptive learning ability, especially with deep configuration. However, their efficacy is inherently limited owing homogenous network formation the unique use of linear convolution. In this study, we propose heterogeneous model which allows greater flexibility embedding additional non-linearity at core data transformation. To end, idea an operational neuron or Operational (ONN),...

Object proposals improve the efficiency of object detection by providing probable locations objects in an image. Most state-of-the-art proposal methods employ a supervised approach and learn features from ground truth annotations. We present novel unsupervised for generating that is based on human visual system quantum mechanical principles. Despite being devoid any learnt priors pertaining to images, proposed method shown yield competitive results with approaches.