-
A Doubly Regularized Linear Discriminant Analysis Classifier with Automatic Parameter Selection
Authors:
Alam Zaib,
Tarig Ballal,
Shahid Khattak,
Tareq Y. Al-Naffouri
Abstract:
Linear discriminant analysis (LDA) based classifiers tend to falter in many practical settings where the training data size is smaller than, or comparable to, the number of features. As a remedy, different regularized LDA (RLDA) methods have been proposed. These methods may still perform poorly depending on the size and quality of the available training data. In particular, the test data deviation…
▽ More
Linear discriminant analysis (LDA) based classifiers tend to falter in many practical settings where the training data size is smaller than, or comparable to, the number of features. As a remedy, different regularized LDA (RLDA) methods have been proposed. These methods may still perform poorly depending on the size and quality of the available training data. In particular, the test data deviation from the training data model, for example, due to noise contamination, can cause severe performance degradation. Moreover, these methods commit further to the Gaussian assumption (upon which LDA is established) to tune their regularization parameters, which may compromise accuracy when dealing with real data. To address these issues, we propose a doubly regularized LDA classifier that we denote as R2LDA. In the proposed R2LDA approach, the RLDA score function is converted into an inner product of two vectors. By substituting the expressions of the regularized estimators of these vectors, we obtain the R2LDA score function that involves two regularization parameters. To set the values of these parameters, we adopt three existing regularization techniques; the constrained perturbation regularization approach (COPRA), the bounded perturbation regularization (BPR) algorithm, and the generalized cross-validation (GCV) method. These methods are used to tune the regularization parameters based on linear estimation models, with the sample covariance matrix's square root being the linear operator. Results obtained from both synthetic and real data demonstrate the consistency and effectiveness of the proposed R2LDA approach, especially in scenarios involving test data contaminated with noise that is not observed during the training phase.
△ Less
Submitted 27 March, 2021; v1 submitted 28 April, 2020;
originally announced April 2020.
-
Deep Learning for Medical Image Processing: Overview, Challenges and Future
Authors:
Muhammad Imran Razzak,
Saeeda Naz,
Ahmad Zaib
Abstract:
Healthcare sector is totally different from other industry. It is on high priority sector and people expect highest level of care and services regardless of cost. It did not achieve social expectation even though it consume huge percentage of budget. Mostly the interpretations of medical data is being done by medical expert. In terms of image interpretation by human expert, it is quite limited due…
▽ More
Healthcare sector is totally different from other industry. It is on high priority sector and people expect highest level of care and services regardless of cost. It did not achieve social expectation even though it consume huge percentage of budget. Mostly the interpretations of medical data is being done by medical expert. In terms of image interpretation by human expert, it is quite limited due to its subjectivity, the complexity of the image, extensive variations exist across different interpreters, and fatigue. After the success of deep learning in other real world application, it is also providing exciting solutions with good accuracy for medical imaging and is seen as a key method for future applications in health secotr. In this chapter, we discussed state of the art deep learning architecture and its optimization used for medical image segmentation and classification. In the last section, we have discussed the challenges deep learning based methods for medical imaging and open research issue.
△ Less
Submitted 22 April, 2017;
originally announced April 2017.
-
Distributed Channel Estimation and Pilot Contamination Analysis for Massive MIMO-OFDM Systems
Authors:
Alam Zaib,
Mudassir Masood,
Anum Ali,
Weiyu Xu,
Tareq Y. Al-Naffouri
Abstract:
Massive MIMO communication systems, by virtue of utilizing very large number of antennas, have a potential to yield higher spectral and energy efficiency in comparison with the conventional MIMO systems. In this paper, we consider uplink channel estimation in massive MIMO-OFDM systems with frequency selective channels. With increased number of antennas, the channel estimation problem becomes very…
▽ More
Massive MIMO communication systems, by virtue of utilizing very large number of antennas, have a potential to yield higher spectral and energy efficiency in comparison with the conventional MIMO systems. In this paper, we consider uplink channel estimation in massive MIMO-OFDM systems with frequency selective channels. With increased number of antennas, the channel estimation problem becomes very challenging as exceptionally large number of channel parameters have to be estimated. We propose an efficient distributed linear minimum mean square error (LMMSE) algorithm that can achieve near optimal channel estimates at very low complexity by exploiting the strong spatial correlations and symmetry of large antenna array elements. The proposed method involves solving a (fixed) reduced dimensional LMMSE problem at each antenna followed by a repetitive sharing of information through collaboration among neighboring antenna elements. To further enhance the channel estimates and/or reduce the number of reserved pilot tones, we propose a data-aided estimation technique that relies on finding a set of most reliable data carriers. We also analyse the effect of pilot contamination on the mean square error (MSE) performance of different channel estimation techniques. Unlike the conventional approaches, we use stochastic geometry to obtain analytical expression for interference variance (or power) across OFDM frequency tones and use it to derive the MSE expressions for different algorithms under both noise and pilot contaminated regimes. Simulation results validate our analysis and the near optimal MSE performance of proposed estimation algorithms.
△ Less
Submitted 29 July, 2015;
originally announced July 2015.
-
Optimal Non-coherent Data Detection for Massive SIMO Wireless Systems with General Constellations: A Polynomial Complexity Solution
Authors:
Haider Ali Jasim Alshamary,
Md Fahim Anjum,
Tareq Al-Naffouri,
Alam Zaib,
Weiyu Xu
Abstract:
Massive MIMO systems can greatly increase spectral and energy efficiency over traditional MIMO systems by exploiting large antenna arrays. However, increasing the number of antennas at the base station (BS) makes the uplink non-coherent data detection very challenging in massive MIMO systems. In this paper we consider the joint maximum likelihood (ML) channel estimation and data detection problem…
▽ More
Massive MIMO systems can greatly increase spectral and energy efficiency over traditional MIMO systems by exploiting large antenna arrays. However, increasing the number of antennas at the base station (BS) makes the uplink non-coherent data detection very challenging in massive MIMO systems. In this paper we consider the joint maximum likelihood (ML) channel estimation and data detection problem for massive SIMO (single input multiple output) wireless systems, which is a special case of wireless systems with large antenna arrays. We propose exact ML non-coherent data detection algorithms for both constant-modulus and nonconstant-modulus constellations, with a low expected complexity. Despite the large number of unknown channel coefficients for massive SIMO systems, we show that the expected computational complexity of these algorithms is linear in the number of receive antennas and polynomial in channel coherence time. Simulation results show the performance gains (up to 5 dB improvement) of the optimal non-coherent data detection with a low computational complexity.
△ Less
Submitted 10 July, 2015; v1 submitted 8 July, 2015;
originally announced July 2015.
-
Optimal non-coherent data detection for massive SIMO wireless systems: A polynomial complexity solution
Authors:
Haider Ali Jasim Alshamary,
Tareq Al-Naffouri,
Alam Zaib,
Weiyu Xu
Abstract:
Massive MIMO systems have made significant progress in increasing spectral and energy efficiency over traditional MIMO systems by exploiting large antenna arrays. In this paper we consider the joint maximum likelihood (ML) channel estimation and data detection problem for massive SIMO (single input multiple output) wireless systems. Despite the large number of unknown channel coefficients for mass…
▽ More
Massive MIMO systems have made significant progress in increasing spectral and energy efficiency over traditional MIMO systems by exploiting large antenna arrays. In this paper we consider the joint maximum likelihood (ML) channel estimation and data detection problem for massive SIMO (single input multiple output) wireless systems. Despite the large number of unknown channel coefficients for massive SIMO systems, we improve an algorithm to achieve the exact ML non-coherent data detection with a low expected complexity. We show that the expected computational complexity of this algorithm is linear in the number of receive antennas and polynomial in channel coherence time. Simulation results show the performance gain of the optimal non-coherent data detection with a low computational complexity.
△ Less
Submitted 25 November, 2014;
originally announced November 2014.
-
A Layered Modeling and Simulation Approach to investigate Resource-aware Computing in MPSoCs
Authors:
Aurang Zaib,
Prashanth Raju,
Thomas Wild,
Andreas Herkersdorf
Abstract:
Increasing complexity of modern multi-processor system on chip (MPSoC) and the decreasing feature size have introduced new challenges. System designers have to consider now aspects which were not part of the design process in past times. Resource-aware Computing is one of such emerging design concerns which can help to improve performance, dependability and resource utilization of overall system.…
▽ More
Increasing complexity of modern multi-processor system on chip (MPSoC) and the decreasing feature size have introduced new challenges. System designers have to consider now aspects which were not part of the design process in past times. Resource-aware Computing is one of such emerging design concerns which can help to improve performance, dependability and resource utilization of overall system. Resource-aware execution takes into account the resource status when executing tasks on MPSoCs. Exploration of resource-aware computing at early design stages of complex systems is mandatory and appropriate methodologies to do this in an efficient manner are thus required. In this paper, we present a modular approach which provides modeling and simulation support for investigation of resource-aware execution in MPSoCs. The proposed methodology enables rapid exploration of the design space by modeling and simulating the resource-awareness in a separate layer while widely reusing the legacy system model in the other layer. Our experiments illustrate the benefits of our approach for the exploration of resource-aware execution on MPSoCs.
△ Less
Submitted 12 May, 2014;
originally announced May 2014.
