Traditional speech enhancement methods often rely on complex signal processing algorithms, which may not be efficient for real-time applications or suffer from limitations in handling various types of noise. Deploying Deep Neural Network (DNN) models resource-constrained environments can challenging due to their high computational requirements. In this paper, we propose a Knowledge Distillation (KD) method leveraging the information stored intermediate latent features very DNN (teacher)...

Despite remarkable success in a variety of computer vision applications, it is well-known that deep learning can fail catastrophically when presented with out-of-distribution data, where there are usually style differences between the training and test images. Toward addressing this challenge, we consider domain generalization problem, wherein predictors trained using data drawn from family related (source) domains then evaluated on distinct unseen domain. Naively model aggregate set (pooled...

Sleep apnea is the most popular sleep disorders which may lead to physical and mental problems. A quick accurate diagnosis helps physicians make a suitable remedy for it. Electroencephalogram (EEG) electrical activity recorded from surface of skull. The identity EEG non-linear complex, thus study complexity signal can be helpful access valuable information In this paper, 12 entropies (Shannon, Renyi, Tsallis, threshold, permutation, spectral, wavelet, SURE, norm, log energy, fuzzy, sample),...

Hashing-based approaches have gained popularity for large-scale image retrieval in recent years. It has been shown that semi-supervised hashing, which incorporates similarity/dissimilarity information into hash function learning could improve the hashing quality. In this paper, we present a novel kernel-based binary model by taking account auxiliary information, i.e., similar and dissimilar data pairs achieving high quality hashing. The main idea is to map points highly non-linear feature...

Recent deep neural network (DNN) models have achieved high performance in speech enhancement. However, deploying such complex resource-constrained environments can be challenging without significant degradation. Knowledge distillation (KD), a technique where smaller (student) model is trained to mimic the behavior of larger, more (teacher) model, has emerged as popular approach address this challenge. In paper, we propose feature-augmentation based knowledge method for enhancement,...

In this paper, taking the advantage of multiple source domains, we propose a novel approach for visual Domain Generalization (DG). The three key ideas underlying our formulation are (1) leveraging disentangled representations images to define different factors variations, (2) generating perturbed by changing such composing images, (3) enforcing learner (classifier) be invariant changes in images. We demonstrate effectiveness on several widely used datasets domain generalization problem, all...

Deep neural networks (DNNs) can achieve high accuracy when there is abundant training data that has the same distribution as test data. In practical applications, deficiency often a concern. For classification tasks, lack of enough labeled images in set results overfitting. Another issue mismatch between and domains, which poor model performance. This calls for need to have robust efficient deep learning models. this work, we propose approach called Multi-Expert Adversarial Regularization...