Farzan Haddadi
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
Farzan Haddadi has a total of 49 co-authors. Here are the top co-authors by number of shared publications:
- Sajad Daei
KTH Royal Institute of Technology, Stockholm, Sweden
10 shared publications - Arash Amini
Electrical Engineering Department, Sharif University of Technology, Tehran, Iran
8 shared publications - Pourya Behmandpoor
Vrije Universiteit Brussel
3 shared publications - Bahman Abolhassani
School of Electrical Engineering, Iran University of Science and Technology (IUST), Tehran, Iran
2 shared publications - Mohammad Reza Aref
Department of Electrical Engineering Sharif University of Technology Tehran Iran
2 shared publications - Mohammad Reza Mehrabani
School of Electrical Engineering, Iran University of Science and Technology, Tehran, Iran
2 shared publications
Latest Publications
- Open Access
Hybrid Quantum-Classical Selective State Space Artificial Intelligence
Hybrid Quantum Classical (HQC) algorithms constitute one of the most effective paradigms for exploiting the computational advantages of quantum systems in large-scale numerical tasks. By operating in high-dimensional Hilbert spaces, quantum circuits enable exponential speed-ups and provide access to richer representations of cost landscapes compared to purely classical methods. These capabilities are particularly relevant for machine learning, where state-of-the-art models especially in Natural Language Processing (NLP) suffer from prohibitive time complexity due to massive matrix multiplications and high-dimensional optimization. In this manuscript, we propose a Hybrid Quantum Classical selection mechanism for the Mamba architecture, designed specifically for temporal sequence classification problems. Our approach leverages Variational Quantum Circuits (VQCs) as quantum gating modules that both enhance feature extraction and improve suppression of irrelevant information. This integration directly addresses the computational bottlenecks of deep learning architectures by exploiting quantum resources for more efficient representation learning. We analyze how introducing quantum subroutines into large language models (LLMs) impacts their generalization capability, expressivity, and parameter efficiency. The results highlight the potential of quantum-enhanced gating mechanisms as a path toward scalable, resource-efficient NLP models, in a limited simulation step. Within the first four epochs on a reshaped MNIST dataset with input format (batch, 784, d_model), our hybrid model achieved 24.6% accuracy while using one quantum layer and achieve higher expressivity, compared to 21.6% obtained by a purely classical selection mechanism. we state No founding
Open AccessHybrid Quantum-Classical Selective State Space Artificial Intelligence
Hybrid Quantum Classical (HQC) algorithms constitute one of the most effective paradigms for exploiting the computational advantages of quantum systems in large-scale numerical tasks. By operating in high-dimensional Hilbert spaces, quantum circuits enable exponential speed-ups and provide access to richer representations of cost landscapes compared to purely classical methods. These capabilities are particularly relevant for machine learning, where state-of-the-art models especially in Natural Language Processing (NLP) suffer from prohibitive time complexity due to massive matrix multiplications and high-dimensional optimization. In this manuscript, we propose a Hybrid Quantum Classical selection mechanism for the Mamba architecture, designed specifically for temporal sequence classification problems. Our approach leverages Variational Quantum Circuits (VQCs) as quantum gating modules that both enhance feature extraction and improve suppression of irrelevant information. This integration directly addresses the computational bottlenecks of deep learning architectures by exploiting quantum resources for more efficient representation learning. We analyze how introducing quantum subroutines into large language models (LLMs) impacts their generalization capability, expressivity, and parameter efficiency. The results highlight the potential of quantum-enhanced gating mechanisms as a path toward scalable, resource-efficient NLP models, in a limited simulation step. Within the first four epochs on a reshaped MNIST dataset with input format (batch, 784, d_model), our hybrid model achieved 24.6% accuracy while using one quantum layer and achieve higher expressivity, compared to 21.6% obtained by a purely classical selection mechanism. we state No founding
- Open Access
- 9 August 2024
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Access
- Vol. 12, 111565-111578
Semi-Blind Channel Estimation and Pilot Allocation for Doubly-Selective Massive MIMO-OFDM Systems
This article addresses the problem of channel estimation and pilot allocation in massive multiple-input multiple-output orthogonal frequency division multiplexing (MIMO-OFDM) networks for sparse doubly selective (DS) channels. In DS channels, due to significant Doppler shifts, channels vary fast, a huge number of coefficients must be estimated, and inter-carrier interference (ICI) must be overcome, which make the problem very challenging. To address these challenges, we reframe the issue as a distributed compressive sensing (DCS) coefficient estimation problem within the basis expansion model (BEM) for channel representation. This reduces the number of channel coefficients significantly as in the new problem, the jointly sparse BEM coefficients are estimated. To further improve estimation performance, we first employ a genetic algorithm to optimize pilot locations based on the mutual coherence of the measurement matrix as the criterion. Secondly, taking the advantage of the complex exponential (CE)-BEM, the correlation of the received signal is used to estimate the channel’s most significant lags (MSLs). Using optimized pilot locations and MSLs, we propose a novel algorithm, called semi-blind block orthogonal matching pursuit (SB-BOMP) for channel estimation. Extensive simulation results confirm that the proposed algorithm outperforms the traditional ones over DS channels.
- 9 August 2024
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Access
- Vol. 12, 111565-111578
Open AccessSemi-Blind Channel Estimation and Pilot Allocation for Doubly-Selective Massive MIMO-OFDM Systems
This article addresses the problem of channel estimation and pilot allocation in massive multiple-input multiple-output orthogonal frequency division multiplexing (MIMO-OFDM) networks for sparse doubly selective (DS) channels. In DS channels, due to significant Doppler shifts, channels vary fast, a huge number of coefficients must be estimated, and inter-carrier interference (ICI) must be overcome, which make the problem very challenging. To address these challenges, we reframe the issue as a distributed compressive sensing (DCS) coefficient estimation problem within the basis expansion model (BEM) for channel representation. This reduces the number of channel coefficients significantly as in the new problem, the jointly sparse BEM coefficients are estimated. To further improve estimation performance, we first employ a genetic algorithm to optimize pilot locations based on the mutual coherence of the measurement matrix as the criterion. Secondly, taking the advantage of the complex exponential (CE)-BEM, the correlation of the received signal is used to estimate the channel’s most significant lags (MSLs). Using optimized pilot locations and MSLs, we propose a novel algorithm, called semi-blind block orthogonal matching pursuit (SB-BOMP) for channel estimation. Extensive simulation results confirm that the proposed algorithm outperforms the traditional ones over DS channels.
- Open Access
Matrix Completion via Nonsmooth Regularization of Fully Connected Neural Networks
Conventional matrix completion methods approximate the missing values by assuming the matrix to be low-rank, which leads to a linear approximation of missing values. It has been shown that enhanced performance could be attained by using nonlinear estimators such as deep neural networks. Deep fully connected neural networks (FCNNs), one of the most suitable architectures for matrix completion, suffer from over-fitting due to their high capacity, which leads to low generalizability. In this paper, we control over-fitting by regularizing the FCNN model in terms of the $\ell_{1}$ norm of intermediate representations and nuclear norm of weight matrices. As such, the resulting regularized objective function becomes nonsmooth and nonconvex, i.e., existing gradient-based methods cannot be applied to our model. We propose a variant of the proximal gradient method and investigate its convergence to a critical point. In the initial epochs of FCNN training, the regularization terms are ignored, and through epochs, the effect of that increases. The gradual addition of nonsmooth regularization terms is the main reason for the better performance of the deep neural network with nonsmooth regularization terms (DNN-NSR) algorithm. Our simulations indicate the superiority of the proposed algorithm in comparison with existing linear and nonlinear algorithms.Open AccessMatrix Completion via Nonsmooth Regularization of Fully Connected Neural Networks
Conventional matrix completion methods approximate the missing values by assuming the matrix to be low-rank, which leads to a linear approximation of missing values. It has been shown that enhanced performance could be attained by using nonlinear estimators such as deep neural networks. Deep fully connected neural networks (FCNNs), one of the most suitable architectures for matrix completion, suffer from over-fitting due to their high capacity, which leads to low generalizability. In this paper, we control over-fitting by regularizing the FCNN model in terms of the $\ell_{1}$ norm of intermediate representations and nuclear norm of weight matrices. As such, the resulting regularized objective function becomes nonsmooth and nonconvex, i.e., existing gradient-based methods cannot be applied to our model. We propose a variant of the proximal gradient method and investigate its convergence to a critical point. In the initial epochs of FCNN training, the regularization terms are ignored, and through epochs, the effect of that increases. The gradual addition of nonsmooth regularization terms is the main reason for the better performance of the deep neural network with nonsmooth regularization terms (DNN-NSR) algorithm. Our simulations indicate the superiority of the proposed algorithm in comparison with existing linear and nonlinear algorithms. - Open Access
- 22 February 2023
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Access
- Vol. 11, 21921-21933
Second-Order Statistics-Aided Channel Estimation for Multipath Massive MIMO-OFDM Systems
This paper develops an efficient channel estimation algorithm based on second-order statistics of time division duplex (TDD) multiuser massive multiple-input multiple-output (MIMO) systems. The algorithm uses the received signal correlation to determine the most significant lags (MSLs) of the received signal. We first employ these MSLs to propose a novel set containing the channel’s four most significant taps (MSTs). Then, by using them, we propose an efficient semi-blind iterative algorithm called enhanced modified-subspace pursuit (EM-SP). It uses the set mentioned above and two theoretical results ( Lemma 1 and Theorem 1 ) to estimate an arbitrary number of MSTs efficiently. Simulation results show that the normalized mean square error (NMSE) of the proposed EM-SP algorithm is much smaller than that of the subspace pursuit (SP) and orthogonal matching pursuit (OMP) algorithms at the cost of 0.3 % and 2 % more computational complexity for the channels with three and six nonzero paths, respectively. Moreover, the NMSE of it is very close to that of the optimal genie-aided least square algorithm.
- 22 February 2023
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Access
- Vol. 11, 21921-21933
Open AccessSecond-Order Statistics-Aided Channel Estimation for Multipath Massive MIMO-OFDM Systems
This paper develops an efficient channel estimation algorithm based on second-order statistics of time division duplex (TDD) multiuser massive multiple-input multiple-output (MIMO) systems. The algorithm uses the received signal correlation to determine the most significant lags (MSLs) of the received signal. We first employ these MSLs to propose a novel set containing the channel’s four most significant taps (MSTs). Then, by using them, we propose an efficient semi-blind iterative algorithm called enhanced modified-subspace pursuit (EM-SP). It uses the set mentioned above and two theoretical results ( Lemma 1 and Theorem 1 ) to estimate an arbitrary number of MSTs efficiently. Simulation results show that the normalized mean square error (NMSE) of the proposed EM-SP algorithm is much smaller than that of the subspace pursuit (SP) and orthogonal matching pursuit (OMP) algorithms at the cost of 0.3 % and 2 % more computational complexity for the channels with three and six nonzero paths, respectively. Moreover, the NMSE of it is very close to that of the optimal genie-aided least square algorithm.
- Open Access
- 18 June 2022
- Published by Springer Nature in EURASIP Journal on Advances in Signal Processing
- Vol. 2022 (1) , 1-13
Adaptive recovery of dictionary-sparse signals using binary measurements
- 18 June 2022
- Published by Springer Nature in EURASIP Journal on Advances in Signal Processing
- Vol. 2022 (1) , 1-13
Open AccessAdaptive recovery of dictionary-sparse signals using binary measurements
- 25 April 2022
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Signal Processing
- Vol. 70 (1053587X) , 3000-3010
Multi-Weight Nuclear Norm Minimization for Low-Rank Matrix Recovery in Presence of Subspace Prior Information
Weighted nuclear norm minimization has been recently recognized as a technique for reconstruction of a low-rank matrix from compressively sampled measurements when some prior information about the column and row subspaces of the matrix is available. We derive the conditions and the associated recovery guarantees of weighted nuclear norm minimization when multiple weights are allowed. This setup could be used when one has access to prior subspaces forming multiple angles with the column and row subspaces of the ground-truth matrix. While existing works in this field use a single weight to penalize all the angles, we propose a multi-weight problem which is designed to penalize each angle independently using a distinct weight. Specifically, we prove that our proposed multi-weight problem is stable and robust under weaker conditions for the measurement operator than the analogous conditions for single-weight scenario and standard nuclear norm minimization. Moreover, it provides better reconstruction error than the state of the art methods. We illustrate our results with extensive numerical experiments that demonstrate the advantages of allowing multiple weights in the recovery procedure. Our work has beneficial implications for channel estimation in multiple-input multiple output (MIMO) wireless communications based on Frequency Division Duplexing (FDD). The existing methods for channel estimation in this application require a huge number of pilot (training) signals to estimate the downlink channel which greatly wastes the spectrum resources in massive MIMO systems. We provide a dynamic channel estimation scenario for FDD massive MIMO systems and show how our method could be applied to enhance the spectral efficiency.
- 25 April 2022
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Signal Processing
- Vol. 70 (1053587X) , 3000-3010
Multi-Weight Nuclear Norm Minimization for Low-Rank Matrix Recovery in Presence of Subspace Prior Information
Weighted nuclear norm minimization has been recently recognized as a technique for reconstruction of a low-rank matrix from compressively sampled measurements when some prior information about the column and row subspaces of the matrix is available. We derive the conditions and the associated recovery guarantees of weighted nuclear norm minimization when multiple weights are allowed. This setup could be used when one has access to prior subspaces forming multiple angles with the column and row subspaces of the ground-truth matrix. While existing works in this field use a single weight to penalize all the angles, we propose a multi-weight problem which is designed to penalize each angle independently using a distinct weight. Specifically, we prove that our proposed multi-weight problem is stable and robust under weaker conditions for the measurement operator than the analogous conditions for single-weight scenario and standard nuclear norm minimization. Moreover, it provides better reconstruction error than the state of the art methods. We illustrate our results with extensive numerical experiments that demonstrate the advantages of allowing multiple weights in the recovery procedure. Our work has beneficial implications for channel estimation in multiple-input multiple output (MIMO) wireless communications based on Frequency Division Duplexing (FDD). The existing methods for channel estimation in this application require a huge number of pilot (training) signals to estimate the downlink channel which greatly wastes the spectrum resources in massive MIMO systems. We provide a dynamic channel estimation scenario for FDD massive MIMO systems and show how our method could be applied to enhance the spectral efficiency.
- 1 June 2021
- Published by Elsevier in Signal Processing
- Vol. 183
Off-the-grid recovery of time and frequency shifts with multiple measurement vectors
- 1 June 2021
- Published by Elsevier in Signal Processing
- Vol. 183
Off-the-grid recovery of time and frequency shifts with multiple measurement vectors
- 19 November 2020
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Signal Processing Letters
- Vol. 28 (10709908) , 583-587
Blind Two-Dimensional Super-Resolution in Multiple-Input Single-Output Linear Systems
In this paper, we consider a multiple-input single-output (MISO) linear time-varying system whose output is a superposition of scaled and time-frequency shifted versions of inputs. The goal of this paper is to determine system characteristics and input signals from the single output signal. More precisely, we want to recover the continuous time-frequency shift pairs, the corresponding (complex-valued) amplitudes and the input signals from only one output vector. This problem arises in a variety of applications such as radar imaging, microscopy, channel estimation and localization problems. While this problem is naturally ill-posed, by constraining the unknown input waveforms to lie in separate known low-dimensional subspaces, it becomes tractable. More explicitly, we propose a semidefinite program which exactly recovers time-frequency shift pairs and input signals. We prove uniqueness and optimality of the solution to this program. Moreover, we provide a grid-based approach which can significantly reduce computational complexity in exchange for adding a small gridding error. Numerical results confirm the ability of our proposed method to exactly recover the unknowns.
- 19 November 2020
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Signal Processing Letters
- Vol. 28 (10709908) , 583-587
Blind Two-Dimensional Super-Resolution in Multiple-Input Single-Output Linear Systems
In this paper, we consider a multiple-input single-output (MISO) linear time-varying system whose output is a superposition of scaled and time-frequency shifted versions of inputs. The goal of this paper is to determine system characteristics and input signals from the single output signal. More precisely, we want to recover the continuous time-frequency shift pairs, the corresponding (complex-valued) amplitudes and the input signals from only one output vector. This problem arises in a variety of applications such as radar imaging, microscopy, channel estimation and localization problems. While this problem is naturally ill-posed, by constraining the unknown input waveforms to lie in separate known low-dimensional subspaces, it becomes tractable. More explicitly, we propose a semidefinite program which exactly recovers time-frequency shift pairs and input signals. We prove uniqueness and optimality of the solution to this program. Moreover, we provide a grid-based approach which can significantly reduce computational complexity in exchange for adding a small gridding error. Numerical results confirm the ability of our proposed method to exactly recover the unknowns.
- 18 November 2020
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Information Theory
- Vol. 67 (2) , 1069-1079
Eigenvectors of Deformed Wigner Random Matrices
We investigate eigenvectors of rank-one deformations of random matrices B = A + θuu* in which A ∈ RN×N is a Wigner real symmetric random matrix, θ ∈ R+, and u is uniformly distributed on the unit sphere. It is well known that for θ > 1 the eigenvector associated with the largest eigenvalue of B closely estimates u asymptotically, while for θ < 1 the eigenvectors of B are uninformative about u. We examine O(1/N) correlation of eigenvectors with u before phase transition and show that eigenvectors with larger eigenvalue exhibit stronger alignment with deforming vector through an explicit inverse law 1/θ*-x with θ* := θ + 1/θ. This distribution function will be shown to be the ordinary generating function of Chebyshev polynomials of the second kind. These polynomials form an orthogonal set with respect to the semicircle weighting function. This law is an increasing function in the support of semicircle law for eigenvalues (-2, +2). Therefore, most of energy of the unknown deforming vector is concentrated in a cN-dimensional (c < 1) known subspace of B. We use a combinatorial approach to prove the result. We also extend the result to constant rank-r deformations.
- 18 November 2020
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Information Theory
- Vol. 67 (2) , 1069-1079
Eigenvectors of Deformed Wigner Random Matrices
We investigate eigenvectors of rank-one deformations of random matrices B = A + θuu* in which A ∈ RN×N is a Wigner real symmetric random matrix, θ ∈ R+, and u is uniformly distributed on the unit sphere. It is well known that for θ > 1 the eigenvector associated with the largest eigenvalue of B closely estimates u asymptotically, while for θ < 1 the eigenvectors of B are uninformative about u. We examine O(1/N) correlation of eigenvectors with u before phase transition and show that eigenvectors with larger eigenvalue exhibit stronger alignment with deforming vector through an explicit inverse law 1/θ*-x with θ* := θ + 1/θ. This distribution function will be shown to be the ordinary generating function of Chebyshev polynomials of the second kind. These polynomials form an orthogonal set with respect to the semicircle weighting function. This law is an increasing function in the support of semicircle law for eigenvalues (-2, +2). Therefore, most of energy of the unknown deforming vector is concentrated in a cN-dimensional (c < 1) known subspace of B. We use a combinatorial approach to prove the result. We also extend the result to constant rank-r deformations.
- 14 January 2020
- Published by Springer Nature in Multidimensional Systems and Signal Processing
- Vol. 31 (3) , 1029-1049
Blind three dimensional deconvolution via convex optimization
- 14 January 2020
- Published by Springer Nature in Multidimensional Systems and Signal Processing
- Vol. 31 (3) , 1029-1049
Blind three dimensional deconvolution via convex optimization