Shaikh Anowarul Fattah
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Previous Organizations / Institutes
The organizations / institutes from the scholar's other articles in the past ten years
Shaikh Anowarul Fattah has a total of 144 co-authors. Here are the top co-authors by number of shared publications:
- Mohammad Saquib
Department of Electrical and Computer Engineering, The University of Texas at Dallas, Richardson, TX, USA
17 shared publications Celia Shahnaz
Department of Electrical and Electronic Engineering, Bangladesh University of Engineering and Technology, Dhaka, Bangladesh
11 shared publications- Tanvir Mahmud
Department of Electrical and Electronic Engineering, Bangladesh University of Engineering and Technology, Dhaka, Bangladesh
11 shared publications - Monjur Morshed
Department of Electrical and Electronic Engineering, Bangladesh University of Engineering and Technology, Dhaka, Bangladesh
5 shared publications - Jishnu Mahmud
Department of Computer Science and Engineering, BRAC University, Dhaka, Bangladesh
4 shared publications - Khan Arif Wahid
Department of Electrical and Computer Engineering, University of Saskatchewan, Saskatoon, Canada
4 shared publications
Latest Publications
- Open Access
- 21 September 2026
- Published by Elsevier in SSRN Electronic Journal
Reconstruction and Forecasting Limits of Highly Sparse Electrical Load Time Series Data
Reliable short-term load forecasting is severely compromised when measurement data is highly incomplete, irregular, and non-stationary. This study investigates load forecasting under extreme data sparsity using a real-world campus (Bangladesh University of Engineering and Technology) electrical datasetcharacterized by approximately $62.5\%$ missing hourly observations. To address this severe missingness, we introduce a novel covariate-conditioned diffusion method specifically designed for the accurate interpolation of heavily sparse load profiles. We demonstrate that this generative approach significantly outperforms traditional statistical and adversarial baselines in reconstruction accuracy, which directly translates to superior downstream forecasting performance when training a standard LSTM model (yielding an $11.8\%$ MAPE). Furthermore, beyond empirical improvements, we establish a formal mathematical relationship between the absolute forecasting limit and the intrinsic volatility of the dataset. By deriving the innovation coefficient of variation, we project a theoretical forecasting floor (approximately $12.5\%$ MAPE) that estimates the best possible forecasting accuracy given the dataset's inherent irregularities. Ultimately, this work provides both a robust imputation method for extreme data sparsity and a principled framework for distinguishing correctable imputation errors from irreducible stochastic uncertainty.- 21 September 2026
- Published by Elsevier in SSRN Electronic Journal
Open AccessReconstruction and Forecasting Limits of Highly Sparse Electrical Load Time Series Data
Reliable short-term load forecasting is severely compromised when measurement data is highly incomplete, irregular, and non-stationary. This study investigates load forecasting under extreme data sparsity using a real-world campus (Bangladesh University of Engineering and Technology) electrical datasetcharacterized by approximately $62.5\%$ missing hourly observations. To address this severe missingness, we introduce a novel covariate-conditioned diffusion method specifically designed for the accurate interpolation of heavily sparse load profiles. We demonstrate that this generative approach significantly outperforms traditional statistical and adversarial baselines in reconstruction accuracy, which directly translates to superior downstream forecasting performance when training a standard LSTM model (yielding an $11.8\%$ MAPE). Furthermore, beyond empirical improvements, we establish a formal mathematical relationship between the absolute forecasting limit and the intrinsic volatility of the dataset. By deriving the innovation coefficient of variation, we project a theoretical forecasting floor (approximately $12.5\%$ MAPE) that estimates the best possible forecasting accuracy given the dataset's inherent irregularities. Ultimately, this work provides both a robust imputation method for extreme data sparsity and a principled framework for distinguishing correctable imputation errors from irreducible stochastic uncertainty. - Open Access
- 31 July 2026
- Published by Elsevier in SSRN Electronic Journal
A MULTI GRAPH ATTENTION FUSION FRAMEWORK FOR LEARNING MULTI-SCALE LOAD DYNAMICS IN SHORT-TERM LOAD FORECASTING
Short-term load forecasting (STLF) is essential for stable and smooth functioning of smart grid operation. However, due to the very complex nature of energy demands, forecasting becomes inefficient. While most of the existing graph-based load forecasting models utilize only one static relational topology and perform directly on interconnected geographical load embeddings, majority of the existing decomposition-based models ignore various behavioral interactions among different time scales. In order to address those shortcomings, this paper presents a new framework termed as CEEMDAN-driven Multi-Graph State-Aware Transformer Network (CGSTNet) that integrates multi-scale signal decomposition, multi varied graph representation learning, and state-aware behavioral fusion. Unlike other existing methods, the presented framework constructs complementary graphs after modelling multiple temporal and behavioral relationships simultaneously using the CEEMDAN decomposition and further Transformer encoder. The multi-graph attention encoder is developed to learn different relational features among the load profiles, which are dynamically updated through a new behavioral-graph fusion mechanism that takes into account state information. Finally, BiLSTM decoder is applied for sequential load predictions. This unified representation learning scheme efficiently bridges different graphs with various load behavior dynamics in one end-to-end optimization framework. The results obtained through experiments demonstrate that the proposed framework always performs better than all the baseline models, having an RMSE value of 903.13, MAE of 656.02, and MAPE of 2.27%, leading to almost a 54% reduction in RMSE compared to the best baseline model. The efficiency and contributions of each proposed module is confirmed by ablation study and scenario-based analysis as well as error analysis.
- 31 July 2026
- Published by Elsevier in SSRN Electronic Journal
Open AccessA MULTI GRAPH ATTENTION FUSION FRAMEWORK FOR LEARNING MULTI-SCALE LOAD DYNAMICS IN SHORT-TERM LOAD FORECASTING
Short-term load forecasting (STLF) is essential for stable and smooth functioning of smart grid operation. However, due to the very complex nature of energy demands, forecasting becomes inefficient. While most of the existing graph-based load forecasting models utilize only one static relational topology and perform directly on interconnected geographical load embeddings, majority of the existing decomposition-based models ignore various behavioral interactions among different time scales. In order to address those shortcomings, this paper presents a new framework termed as CEEMDAN-driven Multi-Graph State-Aware Transformer Network (CGSTNet) that integrates multi-scale signal decomposition, multi varied graph representation learning, and state-aware behavioral fusion. Unlike other existing methods, the presented framework constructs complementary graphs after modelling multiple temporal and behavioral relationships simultaneously using the CEEMDAN decomposition and further Transformer encoder. The multi-graph attention encoder is developed to learn different relational features among the load profiles, which are dynamically updated through a new behavioral-graph fusion mechanism that takes into account state information. Finally, BiLSTM decoder is applied for sequential load predictions. This unified representation learning scheme efficiently bridges different graphs with various load behavior dynamics in one end-to-end optimization framework. The results obtained through experiments demonstrate that the proposed framework always performs better than all the baseline models, having an RMSE value of 903.13, MAE of 656.02, and MAPE of 2.27%, leading to almost a 54% reduction in RMSE compared to the best baseline model. The efficiency and contributions of each proposed module is confirmed by ablation study and scenario-based analysis as well as error analysis.
- 20 July 2026
- Published by Springer Nature in Quantum Machine Intelligence
- Vol. 8  (2) , 79
Selective feature re-encoded quantum convolutional neural network with joint optimization for image classification
- 20 July 2026
- Published by Springer Nature in Quantum Machine Intelligence
- Vol. 8  (2) , 79
Selective feature re-encoded quantum convolutional neural network with joint optimization for image classification
- 25 March 2026
- Published by Springer Nature in Circuits, Systems, and Signal Processing
- Vol. 45  (10) , 8253-8280
An Apnea Detection Method Based on Temporal Feature Variational Pattern of Multi-band EEG Signal Incorporating Sleep Stage Information
- 25 March 2026
- Published by Springer Nature in Circuits, Systems, and Signal Processing
- Vol. 45  (10) , 8253-8280
An Apnea Detection Method Based on Temporal Feature Variational Pattern of Multi-band EEG Signal Incorporating Sleep Stage Information
- Open Access
- 1 January 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Access
- Vol. 14, 152512-152527
DecodeMI-Net: A Multi-Perspective Deep Neural Network With Channel-Wise Attention for Motor Imagery EEG Decoding
- 1 January 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Access
- Vol. 14, 152512-152527
Open AccessDecodeMI-Net: A Multi-Perspective Deep Neural Network With Channel-Wise Attention for Motor Imagery EEG Decoding
- Open Access
- 31 December 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Access
- Vol. 14, 4063-4079
A Multi-Level Probabilistic Deep Learning Network Augmented With Normalizing Flow for Ambiguous Medical Image Segmentation
Medical image segmentation often involves inherent uncertainty due to inter observer variability. In this case, a single deterministic mask obtained by conventional segmentation networks, such as U-Net, cannot capture the distribution of plausible expert annotations, risking missed clinically relevant variants. In order to enable uncertainty quantification and reflect inter expert variability, probabilistic models like Probabilistic U-Net are used, which perform aleatoric or ambiguous segmentation where a latent space is sampled to generate multiple segmentation masks. However, the common use of a conditioned unimodal posterior in these models fails to represent true multimodality, leading to mode bias and limited diversity. To address these limitations, a multi-level Probabilistic U-Net augmented with normalizing flows is proposed to enhance the expressiveness of the latent distribution. The multi-level design induces multiple latent distributions in separate levels of U-Net, enabling more diverse sampling, while the flow module transforms the posterior to add data required modes and expand representational capacity, thereby enriching the expressiveness of the distributions. The proposed flow incorporated multi-level network enables a more flexible and powerful distribution, thereby enhancing the model’s ability to generate high fidelity segmentation masks. Extensive experiments on some publicly available datasets with multiple expert annotations per image demonstrate that the proposed model reduces generalized energy distance (GED), preserves clinically meaningful diversity and sharpens boundary fidelity, with latent grid analyses indicating fuller mode coverage and fewer artifacts. Collectively, these results indicate that the proposed framework advances accuracy, robustness, and clinical reliability for aleatoric, uncertainty aware medical image segmentation.
- 31 December 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Access
- Vol. 14, 4063-4079
Open AccessA Multi-Level Probabilistic Deep Learning Network Augmented With Normalizing Flow for Ambiguous Medical Image Segmentation
Medical image segmentation often involves inherent uncertainty due to inter observer variability. In this case, a single deterministic mask obtained by conventional segmentation networks, such as U-Net, cannot capture the distribution of plausible expert annotations, risking missed clinically relevant variants. In order to enable uncertainty quantification and reflect inter expert variability, probabilistic models like Probabilistic U-Net are used, which perform aleatoric or ambiguous segmentation where a latent space is sampled to generate multiple segmentation masks. However, the common use of a conditioned unimodal posterior in these models fails to represent true multimodality, leading to mode bias and limited diversity. To address these limitations, a multi-level Probabilistic U-Net augmented with normalizing flows is proposed to enhance the expressiveness of the latent distribution. The multi-level design induces multiple latent distributions in separate levels of U-Net, enabling more diverse sampling, while the flow module transforms the posterior to add data required modes and expand representational capacity, thereby enriching the expressiveness of the distributions. The proposed flow incorporated multi-level network enables a more flexible and powerful distribution, thereby enhancing the model’s ability to generate high fidelity segmentation masks. Extensive experiments on some publicly available datasets with multiple expert annotations per image demonstrate that the proposed model reduces generalized energy distance (GED), preserves clinically meaningful diversity and sharpens boundary fidelity, with latent grid analyses indicating fuller mode coverage and fewer artifacts. Collectively, these results indicate that the proposed framework advances accuracy, robustness, and clinical reliability for aleatoric, uncertainty aware medical image segmentation.
- Open Access
- 23 September 2025
- Published by Elsevier in Results in Engineering
- Vol. 28
Deep learning based murmur detection from PCG signals collected at four valve locations using joint optimization and decision fusion
- 23 September 2025
- Published by Elsevier in Results in Engineering
- Vol. 28
Open AccessDeep learning based murmur detection from PCG signals collected at four valve locations using joint optimization and decision fusion
- Open Access
Selective Feature Re-Encoded Quantum Convolutional Neural Network with Joint Optimization for Image Classification
Quantum Machine Learning (QML) has seen significant advancements, driven by recent improvements in Noisy Intermediate-Scale Quantum (NISQ) devices. Leveraging quantum principles such as entanglement and superposition, quantum convolutional neural networks (QCNNs) have demonstrated promising results in classifying both quantum and classical data. This study examines QCNNs in the context of image classification and proposes a novel strategy to enhance feature processing and a QCNN architecture for improved classification accuracy. First, a selective feature re-encoding strategy is proposed, which directs the quantum circuits to prioritize the most informative features, thereby effectively navigating the crucial regions of the Hilbert space to find the optimal solution space. Secondly, a novel parallel-mode QCNN architecture is designed to simultaneously incorporate features extracted by two classical methods, Principal Component Analysis (PCA) and Autoencoders, within a unified training scheme. The joint optimization involved in the training process allows the QCNN to benefit from complementary feature representations, enabling better mutual readjustment of model parameters. To assess these methodologies, comprehensive experiments have been performed using the widely used MNIST and Fashion MNIST datasets for binary classification tasks. Experimental findings reveal that the selective feature re-encoding method significantly improves the quantum circuit's feature processing capability and performance. Furthermore, the jointly optimized parallel QCNN architecture consistently outperforms the individual QCNN models and the traditional ensemble approach involving independent learning followed by decision fusion, confirming its superior accuracy and generalization capabilities.
Open AccessSelective Feature Re-Encoded Quantum Convolutional Neural Network with Joint Optimization for Image Classification
Quantum Machine Learning (QML) has seen significant advancements, driven by recent improvements in Noisy Intermediate-Scale Quantum (NISQ) devices. Leveraging quantum principles such as entanglement and superposition, quantum convolutional neural networks (QCNNs) have demonstrated promising results in classifying both quantum and classical data. This study examines QCNNs in the context of image classification and proposes a novel strategy to enhance feature processing and a QCNN architecture for improved classification accuracy. First, a selective feature re-encoding strategy is proposed, which directs the quantum circuits to prioritize the most informative features, thereby effectively navigating the crucial regions of the Hilbert space to find the optimal solution space. Secondly, a novel parallel-mode QCNN architecture is designed to simultaneously incorporate features extracted by two classical methods, Principal Component Analysis (PCA) and Autoencoders, within a unified training scheme. The joint optimization involved in the training process allows the QCNN to benefit from complementary feature representations, enabling better mutual readjustment of model parameters. To assess these methodologies, comprehensive experiments have been performed using the widely used MNIST and Fashion MNIST datasets for binary classification tasks. Experimental findings reveal that the selective feature re-encoding method significantly improves the quantum circuit's feature processing capability and performance. Furthermore, the jointly optimized parallel QCNN architecture consistently outperforms the individual QCNN models and the traditional ensemble approach involving independent learning followed by decision fusion, confirming its superior accuracy and generalization capabilities.
- 19 May 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Geoscience and Remote Sensing
- Vol. 63  (01962892) , 1-15
SSGRAM: 3-D Spectral-Spatial Feature Network Enhanced by Graph Attention Map for Hyperspectral Image Classification
Convolutional Neural Networks (CNN) and Graph Neural Networks (GNN) are two widely used architectures in Hyperspectral Image (HSI) Classification. Most CNN models tend to heavily rely on neighboring pixels of the target pixel, leading to performance degradation when target pixels differ from their surrounding pixels. To address this limitation, this paper introduces a novel approach to Hyperspectral Image (HSI) classification using a combination of CNN and GNN networks to complement the drawbacks of CNN models. We propose a CNN-based 3D-Spectral Spatial Feature Network (3D-S2FN) to extract features in spectral, spatial, and spectral-spatial spaces, utilizing pyramid squeeze attention to prioritize their importance. Additionally, we introduce a Graph Attention Feature Processor (GAFP) module that evaluates the relevance of neighboring HSI pixels to the target pixel, generating a graph attention map (GRAM). This GRAM is applied to the CNN’s intermediate features to mitigate neighborhood bias. Unlike conventional GNNs, the proposed GAFP module directly processes the data within a local window, without requiring dimensionality reduction methods that would diminish pixel-level detail. By integrating features from the GAFP and the attention-enhanced 3D-S2FN, the proposed SSGRAM method achieves superior classification performance, as demonstrated through extensive experiments on four publicly available datasets.
- 19 May 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Geoscience and Remote Sensing
- Vol. 63  (01962892) , 1-15
SSGRAM: 3-D Spectral-Spatial Feature Network Enhanced by Graph Attention Map for Hyperspectral Image Classification
Convolutional Neural Networks (CNN) and Graph Neural Networks (GNN) are two widely used architectures in Hyperspectral Image (HSI) Classification. Most CNN models tend to heavily rely on neighboring pixels of the target pixel, leading to performance degradation when target pixels differ from their surrounding pixels. To address this limitation, this paper introduces a novel approach to Hyperspectral Image (HSI) classification using a combination of CNN and GNN networks to complement the drawbacks of CNN models. We propose a CNN-based 3D-Spectral Spatial Feature Network (3D-S2FN) to extract features in spectral, spatial, and spectral-spatial spaces, utilizing pyramid squeeze attention to prioritize their importance. Additionally, we introduce a Graph Attention Feature Processor (GAFP) module that evaluates the relevance of neighboring HSI pixels to the target pixel, generating a graph attention map (GRAM). This GRAM is applied to the CNN’s intermediate features to mitigate neighborhood bias. Unlike conventional GNNs, the proposed GAFP module directly processes the data within a local window, without requiring dimensionality reduction methods that would diminish pixel-level detail. By integrating features from the GAFP and the attention-enhanced 3D-S2FN, the proposed SSGRAM method achieves superior classification performance, as demonstrated through extensive experiments on four publicly available datasets.