Taha Bouhsine

Mr. Taha Bouhsine


Research Focus & Metrics

Academic Career Length

3Years (2024-Present)

(Calculated once Scilit publication data is synced: current year − year of first publication + 1)

Representative Publications

View All Publications
15 May 2025 in IEEE Communications Letters
Verified
Journal Article
Closed access
AFAtik Faysal
Atik Faysal
Taha BouhsineTaha Bouhsine
Taha Bouhsine
MRMohammad Rostami
Mohammad Rostami
RRReihaneh Gh. Roshan
Reihaneh Gh. Roshan
HWHuaxia Wang
Huaxia Wang
YYYudong Yao
Yudong Yao
We propose Denoising Masked Autoencoder (Deno-MAE), a novel multimodal autoencoder framework for denoising modulation signals during pretraining. DenoMAE extends the concept of masked autoencoders by incorporating multiple input modalities, including noise as an explicit modality, to enhance cross-modal learning and improve denoising performance. The network is pre-trained using unlabeled noisy modulation signals and constellation diagrams, effectively learning to reconstruct their equivalent noiseless signals and diagrams. DenoMAE achieves state-of-the-art accuracy in automatic modulation classification (AMC) tasks with significantly fewer training samples, demonstrating a 10× reduction in unlabeled pretraining data and a 3× reduction in labeled fine-tuning data compared to existing approaches. Moreover, our model exhibits robust performance across varying Signal-to-Noise Ratios (SNRs) and supports extrapolation on unseen lower SNRs. The results indicate that DenoMAE is an efficient, flexible, and data-efficient solution for denoising and classifying modulation signals in challenging, noise-intensive environments. Our codes are public at https: //github.com/atik666/denoMae/tree/master.

28 October 2025 in IEEE Transactions on Communications
Verified
Journal Article
Closed access
AFAtik Faysal
Atik Faysal
MRMohammad Rostami
Mohammad Rostami
Taha BouhsineTaha Bouhsine
Taha Bouhsine
RRReihaneh Gh. Roshan
Reihaneh Gh. Roshan
NMNikhil Muralidhar
Nikhil Muralidhar
HWHuaxia Wang
Huaxia Wang
We introduce DenoMAE2.0, an enhanced denoising masked autoencoder designed to significantly improve representation learning for Automatic Modulation Classification (AMC) in wireless communications. Unlike standard Masked Autoencoders (MAEs), which solely reconstruct masked inputs, DenoMAE2.0 jointly performs denoising and reconstruction by incorporating a position-aware local patch classification objective. This approach enables the model to simultaneously denoise corrupted signals and accurately reconstruct missing information, effectively capturing both global context and local structural patterns crucial for modulation classification under noisy and data-scarce scenarios. Extensive experiments demonstrate that DenoMAE2.0 achieves superior denoising and classification performance, outperforming baseline methods and its predecessor, DenoMAE, particularly under extremely low Signal-to-Noise Ratios (SNR) (down to -7dB), where other approaches significantly degrade. Our method achieves state-of-the-art accuracy of 82.4%, improving upon the original DenoMAE by 1.1%, while consistently enhancing reconstruction quality across all modulation classes. Additionally, DenoMAE2.0 exhibits robust transfer learning capabilities, achieving consistent improved accuracy over a Vision Transformer (ViT) at different SNRs with a 6.13% highest gain at noise-signal equilibrium point. These results highlight the effectiveness of our dual-objective, self-supervised framework for robust AMC in challenging wireless communication environments.

12 August 2024 in Transportation Research Record: Journal of the Transportation Research Board
Verified
Journal Article
Closed access
CHChristian Hecht
Christian Hecht
SSSurya Teja Swarna
Surya Teja Swarna
PBParth Bhavsar
Parth Bhavsar
YMYusuf Mehta
Yusuf Mehta
Taha BouhsineTaha Bouhsine
Taha Bouhsine
One of the most challenging issues for municipalities in the U.S. is to secure federal funding, state funding, or both, for local roadway improvement. Existing frameworks such as manual data collection, light detection and ranging have proven to be expensive and cumbersome. In this paper, a low-cost pavement management framework is proposed using artificial intelligence (AI). AI has solidified itself across industries as a revolutionary advancement that can automate many tasks that were performed by humans. AI has the potential to make roadway assessment easier and more cost-effective than ever, but this application has been hindered by dataset quality and quantity. Roadway datasets are often imbalanced, containing many more images of certain deformations than others. This decreases the performance of AI models. In this paper, different methods of pavement dataset labeling are tested to gain an understanding of which is best for pavement distress detection using a classification neural network. An AI-friendly pavement condition index is designed to give a clear indicator of the current pavement condition and provide a metric by which to rank the roads based on the need to repair them. The best-performing AI model is incorporated into the low-cost pavement management framework.

7 October 2024 in arXiv
Verified
Preprint
Open access
PDF file
Taha BouhsineTaha Bouhsine
Taha Bouhsine
IAImad El Aaroussi
Imad El Aaroussi
AFAtik Faysal
Atik Faysal
WHWang Huaxia
Wang Huaxia
We introduce a novel anchor-free contrastive learning (AFCL) method leveraging our proposed Similarity-Orthogonality (SimO) loss. Our approach minimizes a semi-metric discriminative loss function that simultaneously optimizes two key objectives: reducing the distance and orthogonality between embeddings of similar inputs while maximizing these metrics for dissimilar inputs, facilitating more fine-grained contrastive learning. The AFCL method, powered by SimO loss, creates a fiber bundle topological structure in the embedding space, forming class-specific, internally cohesive yet orthogonal neighborhoods. We validate the efficacy of our method on the CIFAR-10 dataset, providing visualizations that demonstrate the impact of SimO loss on the embedding space. Our results illustrate the formation of distinct, orthogonal class neighborhoods, showcasing the method's ability to create well-structured embeddings that balance class separation with intra-class variability. This work opens new avenues for understanding and leveraging the geometric properties of learned representations in various machine learning tasks.

13 August 2025 in Preprints.org
Verified
Preprint
Open access
PDF file
Taha BouhsineTaha Bouhsine
Taha Bouhsine
The standard dot product, foundational to deep learning, conflates magnitude and direction, limiting geometric expressiveness and often necessitating additional architectural components such as activation and normalization layers. We introduce the \ⵟ-product (Yat-product), a novel neural operator inspired by physical inverse-square laws, which intrinsically unifies vector alignment and spatial proximity within a single, non-linear, and self-regulating computation. This operator forms the basis of Neural-Matter Networks (NMNs), a new class of architectures that embed non-linearity and normalization directly into the core interaction mechanism, obviating the need for separate activation or normalization layers. We demonstrate that NMNs, and their convolutional and attention-based extensions, achieve competitive or superior performance on benchmark tasks in image classification and language modeling, while yielding more interpretable and geometrically faithful representations. Theoretical analysis establishes the \ⵟ-product as a positive semi-definite Mercer kernel with universal approximation and stable gradient properties. Our results suggest a new design paradigm for deep learning: by grounding neural computation in geometric and physical principles, we can build models that are not only efficient and robust, but also inherently interpretable.