Ashish James - Academia.edu (original) (raw)

Papers by Ashish James

Data detection in magnetic recording (MR) channels can be viewed as an image processing problem, ... more Data detection in magnetic recording (MR) channels can be viewed as an image processing problem, proceeding from the 2-D image of readback bits, to higher level abstractions of features using convolutional layers that finally allow classification of individual bits. In this work, convolutional neural networks (ConvNets) are employed in place of the typical partial response equalizer and maximum-likelihood detector with noise prediction to directly process the un-equalized readback signals and output soft estimates. Several variations of ConvNets are compared in terms of network complexity and performance. The best performing ConvNet detector with two convolutional layers provides a data storage density of up to 3.7489 Terabits/in2 on a low track pitch two-dimensional MR channel simulated with a grain-flipping-probability (GFP) model. An alternate ConvNet architecture reduces the network complexity by about 74%, yet results in only a 2.09% decrease in density compared to the best performing detector.

IEEE Transactions on Magnetics, Mar 1, 2021

This article presents a concatenated Bahl–Cocke–Jelinek–Raviv (BCJR) detector, low-density parity... more This article presents a concatenated Bahl–Cocke–Jelinek–Raviv (BCJR) detector, low-density parity-check (LDPC) decoder, and deep neural network (DNN) architecture for a turbo-detection system for 1-D and 2-D magnetic recording (1DMR and TDMR). The input readings first are fed to a partial response (PR) equalizer. Two types of the equalizer are investigated: a linear filter equalizer with a 1-D/2-D PR target and a convolutional neural network (CNN) PR equalizer that is proposed in this work. The equalized inputs are passed to the BCJR to generate the log-likelihood-ratio (LLR) outputs. We input the BCJR LLRs to a CNN noise predictor to predict the signal-dependent media noise. Two different CNN interfaces with the channel decoder are evaluated for TDMR. Then, the second pass of the BCJR is provided with the estimated media noise, and it feeds its output to the LDPC decoder. The system exchanges LLRs between BCJR, LDPC, and CNN iteratively to achieve higher areal density. The simulation results are performed on a grain flipping probabilistic (GFP) model with 11.4 Teragrains per square inch (Tg/in2). For the GFP data with 18 nm track pitch (TP) and 11 nm bit length (BL), the proposed method for TDMR achieves 27.78% areal density gain over the 1-D pattern-dependent noise prediction (PDNP). The presented BCJR-LDPC-CNN turbo-detection system obtains 3.877 Terabits per square inch (T/bin2) areal density for 11.4 Tg/in2 GFP model data, which is among the highest areal densities reported to date.

arXiv (Cornell University), Nov 8, 2019

Regular inspection of rail valves and engines is an important task to ensure safety and efficienc... more Regular inspection of rail valves and engines is an important task to ensure safety and efficiency of railway networks around the globe. Over the past decade, computer vision and pattern recognition based techniques have gained traction for such inspection and defect detection tasks. An automated end-to-end trained system can potentially provide a low-cost, high throughput, and cheap alternative to manual visual inspection of these components. However, such systems require huge amount of defective images for networks to understand complex defects. In this paper, a multi-phase deep learning based technique is proposed to perform accurate fault detection of rail-valves. Our approach uses a two-step method to perform high precision image segmentation of rail-valves resulting in pixel-wise accurate segmentation. Thereafter, a computer vision technique is used to identify faulty valves. We demonstrate that the proposed approach results in improved detection performance when compared to current state-of-theart techniques used in fault detection.

International Journal of Artificial Intelligence & Applications

The impressive predictive performance of deep learning techniques on a wide range of tasks has le... more The impressive predictive performance of deep learning techniques on a wide range of tasks has led to its widespread use. Estimating the confidence of these predictions is paramount for improving the safety and reliability of such systems. However, the uncertainty estimates provided by neural networks (NNs) tend to be overconfident and unreasonable. Previous studies have found out that ensemble of NNs typically produce good predictions and uncertainty estimates. Inspired by these, this paper presents a new framework that can quantitatively estimate the uncertainties by leveraging the advances in multi-task learning through slight modification to the existing training pipelines. This promising algorithm is developed with an intention of deployment in real world problems which already boast a good predictive performance by reusing those pretrained models. The idea is to capture the behavior of the trained NNs for the base task by augmenting it with the uncertainty estimates from a sup...

IEEE Transactions on Magnetics

2022 IEEE International Symposium on Circuits and Systems (ISCAS)

2019 32nd IEEE International System-on-Chip Conference (SOCC)

Analog circuits are strictly designed under operational, functional and technology constraints. T... more Analog circuits are strictly designed under operational, functional and technology constraints. Together, these bounds create a sparse multi-dimensional design optimization space with the scarcity of labeled analog training data making supervised learning methods ineffective. Accurate approximation of multi-target analog circuits, therefore, requires generation of labeled data around dominant bias and with relevant variance. With such an approach, we explore state-of-theart semi-supervised, generative adversarial network (GAN) towards analog performance modeling. We report on various multi-target analog circuit classification experiments and demonstrate stable GAN performance achieving 2-5% higher accuracy and utilizing only 10% fully simulated manually annotated labeled data against supervised learning methods.

IEEE Transactions on Magnetics, 2017

Efficient multihop transmission scheme for error-free relay forwarding in cooperative networks

Wireless Communications and Mobile Computing, 2013

Multihop cooperative communication is emerging as a key concept to extend the coverage area of th... more Multihop cooperative communication is emerging as a key concept to extend the coverage area of the network and potentially increase the capacity. The spectral efficiency of such networks can be improved by adapting the transmission to time-varying channel conditions, referred to as incremental relaying. Although such incremental relaying concepts are progressively being studied, many challenges, such as erroneous transmissions by intermediate nodes and end-to-end delay of the network, limit its practical use due to lack of an efficient implementation. This paper proposes an efficient multihop incremental relaying technique. In this method, erroneous relay forwarding is mitigated, and the overhead for coordination among nodes is reduced by exploiting the implicit feedback channel available due to the broadcast nature of wireless transmissions. The proposed scheme fully leverages the benefit of overhearing and eliminates the additional feedback slots required for validation. Further, it ensures reliable forwarding of information, which optimizes the throughput of multihop networks. Thorough analysis of the proposed scheme is performed under different deployment environments, and the theoretical analyses presented in this paper are supported with results from extensive simulation studies. Copyright

Abstract—Cooperation among nodes is proposed as an effective means of combating fading and for en... more Abstract—Cooperation among nodes is proposed as an effective means of combating fading and for enhancing system’s overall capacity and coverage. The robustness of rateless codes makes them particularly attractive for such networks. Yet, the decoding aspects of rateless codes are discarded in most of the traditional distributed networks. In this paper, the intermediate packet decodability of rateless codes is exploited to reduce the number of packets being processed at the relay nodes. This is achieved by harnessing the back channel (from destination to relays) for feedback. The relay transmissions and processing are thereby confined to assist the receiver in decoding the remaining information. It is shown both analytically and through simulation studies that such a scheme achieves significant savings in computation complexity, memory usage and overall energy consumption. I.

IEEE Transactions on Magnetics

2013 IEEE 24th Annual International Symposium on Personal, Indoor, and Mobile Radio Communications (PIMRC), 2013

ABSTRACT

The Joint Viterbi detector decoder (JVDD) has been proposed as an alternative to the iterative de... more The Joint Viterbi detector decoder (JVDD) has been proposed as an alternative to the iterative detector, performing both detection and decoding in two stages on a trellis. The first stage estimates and retains a set of survivors, while the second stage performs a parity check on these to compute the minimum metric legal codeword (MMLC). With this structure, near optimal maximum-likelihood decoding (MLD) performance can be achieved but at the cost of complexity especially at long codeword lengths (CWL). JVDD codes have been introduced with the explicit target of reducing this complexity. Further, lower rate codes with more parity checks leads to reduced number of survivors in the JVDD trellis resulting in lower complexity. However, it has been observed that JVDD code performance degrades at low-rates while operating in the low SNR region through an error-floor. This aspect is analyzed in this paper and this performance can be attributed to the JVDD code structure where a constant num...

2020 IEEE 31st Magnetic Recording Conference (TMRC), 2020

The hard disk drive (HDD) industry is facing a physical limit on the areal density (AD) of one-di... more The hard disk drive (HDD) industry is facing a physical limit on the areal density (AD) of one-dimensional magnetic recording (1DMR) on traditional magnetic media. To increase capacity without media redesign, twodimensional magnetic recording (TDMR) has been introduced. The effective channel model has a media noise term which models signal dependent noise due to, e.g., magnetic grains intersected by bit boundaries. Trellis based detection with pattern dependent noise prediction (PDNP) [1] is standard practice in HDDs. The trellis detector sends soft coded bit estimates to a channel decoder, which outputs user information bit estimates. PDNP uses a relatively simple autoregressive noise model and linear prediction; this model is somewhat restrictive and may not accurately represent the media noise, especially at high storage densities. To address this modeling problem, we design and train deep neural network (DNN) based media noise predictors. As DNN [2] models are more general than autoregressive models, they more accurately model media noise compared to PDNP. The proposed turbo detector assumes a channel model for the k th linear equalizer filter output y(k):

2017 IEEE 85th Vehicular Technology Conference (VTC Spring), 2017

The rapid growth of commercial cellular and personal communication systems has increased the dema... more The rapid growth of commercial cellular and personal communication systems has increased the demand for high data rate, power and bandwidth efficient transmissions. Advanced coding and modulation techniques are suitable candidates for this purpose as the potential to reduce the symbol period is often limited by the multipath-induced intersymbol interference (ISI). In this paper, a scheme that jointly performs detection, demodulation, and decoding is proposed and is coined the joint detector demodulator decoder (JDDD). The JDDD is a generalization of the previously proposed joint Viterbi detector decoder (JVDD) that performs detection and decoding of binary phase shift keying (BPSK) modulated signals over ISI channels. The JVDD algorithm performs two operations on a trellis: the first operation computes metrics and retains a set of survivors, while the second operation performs parity checking on the survivors to return the minimum metric legal codeword (MMLC). The JDDD algorithm also has a similar structure but the number of states in JDDD trellis is a factor of the number of symbols defined by the modulator and the characteristics of the ISI channel. In this paper, the performance of the proposed JDDD algorithm is evaluated over an ISI channel.

2019 IEEE Intelligent Transportation Systems Conference (ITSC), 2019

IEEE Transactions on Magnetics, 2021

Data detection in magnetic recording (MR) channels can be viewed as an image processing problem, ... more Data detection in magnetic recording (MR) channels can be viewed as an image processing problem, proceeding from the 2-D image of readback bits, to higher level abstractions of features using convolutional layers that finally allow classification of individual bits. In this work, convolutional neural networks (ConvNets) are employed in place of the typical partial response equalizer and maximum-likelihood detector with noise prediction to directly process the un-equalized readback signals and output soft estimates. Several variations of ConvNets are compared in terms of network complexity and performance. The best performing ConvNet detector with two convolutional layers provides a data storage density of up to 3.7489 Terabits/in2 on a low track pitch two-dimensional MR channel simulated with a grain-flipping-probability (GFP) model. An alternate ConvNet architecture reduces the network complexity by about 74%, yet results in only a 2.09% decrease in density compared to the best performing detector.

Data detection in magnetic recording (MR) channels can be viewed as an image processing problem, ... more Data detection in magnetic recording (MR) channels can be viewed as an image processing problem, proceeding from the 2-D image of readback bits, to higher level abstractions of features using convolutional layers that finally allow classification of individual bits. In this work, convolutional neural networks (ConvNets) are employed in place of the typical partial response equalizer and maximum-likelihood detector with noise prediction to directly process the un-equalized readback signals and output soft estimates. Several variations of ConvNets are compared in terms of network complexity and performance. The best performing ConvNet detector with two convolutional layers provides a data storage density of up to 3.7489 Terabits/in2 on a low track pitch two-dimensional MR channel simulated with a grain-flipping-probability (GFP) model. An alternate ConvNet architecture reduces the network complexity by about 74%, yet results in only a 2.09% decrease in density compared to the best performing detector.

IEEE Transactions on Magnetics, Mar 1, 2021

This article presents a concatenated Bahl–Cocke–Jelinek–Raviv (BCJR) detector, low-density parity... more This article presents a concatenated Bahl–Cocke–Jelinek–Raviv (BCJR) detector, low-density parity-check (LDPC) decoder, and deep neural network (DNN) architecture for a turbo-detection system for 1-D and 2-D magnetic recording (1DMR and TDMR). The input readings first are fed to a partial response (PR) equalizer. Two types of the equalizer are investigated: a linear filter equalizer with a 1-D/2-D PR target and a convolutional neural network (CNN) PR equalizer that is proposed in this work. The equalized inputs are passed to the BCJR to generate the log-likelihood-ratio (LLR) outputs. We input the BCJR LLRs to a CNN noise predictor to predict the signal-dependent media noise. Two different CNN interfaces with the channel decoder are evaluated for TDMR. Then, the second pass of the BCJR is provided with the estimated media noise, and it feeds its output to the LDPC decoder. The system exchanges LLRs between BCJR, LDPC, and CNN iteratively to achieve higher areal density. The simulation results are performed on a grain flipping probabilistic (GFP) model with 11.4 Teragrains per square inch (Tg/in2). For the GFP data with 18 nm track pitch (TP) and 11 nm bit length (BL), the proposed method for TDMR achieves 27.78% areal density gain over the 1-D pattern-dependent noise prediction (PDNP). The presented BCJR-LDPC-CNN turbo-detection system obtains 3.877 Terabits per square inch (T/bin2) areal density for 11.4 Tg/in2 GFP model data, which is among the highest areal densities reported to date.

arXiv (Cornell University), Nov 8, 2019

Regular inspection of rail valves and engines is an important task to ensure safety and efficienc... more Regular inspection of rail valves and engines is an important task to ensure safety and efficiency of railway networks around the globe. Over the past decade, computer vision and pattern recognition based techniques have gained traction for such inspection and defect detection tasks. An automated end-to-end trained system can potentially provide a low-cost, high throughput, and cheap alternative to manual visual inspection of these components. However, such systems require huge amount of defective images for networks to understand complex defects. In this paper, a multi-phase deep learning based technique is proposed to perform accurate fault detection of rail-valves. Our approach uses a two-step method to perform high precision image segmentation of rail-valves resulting in pixel-wise accurate segmentation. Thereafter, a computer vision technique is used to identify faulty valves. We demonstrate that the proposed approach results in improved detection performance when compared to current state-of-theart techniques used in fault detection.

International Journal of Artificial Intelligence & Applications

The impressive predictive performance of deep learning techniques on a wide range of tasks has le... more The impressive predictive performance of deep learning techniques on a wide range of tasks has led to its widespread use. Estimating the confidence of these predictions is paramount for improving the safety and reliability of such systems. However, the uncertainty estimates provided by neural networks (NNs) tend to be overconfident and unreasonable. Previous studies have found out that ensemble of NNs typically produce good predictions and uncertainty estimates. Inspired by these, this paper presents a new framework that can quantitatively estimate the uncertainties by leveraging the advances in multi-task learning through slight modification to the existing training pipelines. This promising algorithm is developed with an intention of deployment in real world problems which already boast a good predictive performance by reusing those pretrained models. The idea is to capture the behavior of the trained NNs for the base task by augmenting it with the uncertainty estimates from a sup...

IEEE Transactions on Magnetics

2022 IEEE International Symposium on Circuits and Systems (ISCAS)

2019 32nd IEEE International System-on-Chip Conference (SOCC)

Analog circuits are strictly designed under operational, functional and technology constraints. T... more Analog circuits are strictly designed under operational, functional and technology constraints. Together, these bounds create a sparse multi-dimensional design optimization space with the scarcity of labeled analog training data making supervised learning methods ineffective. Accurate approximation of multi-target analog circuits, therefore, requires generation of labeled data around dominant bias and with relevant variance. With such an approach, we explore state-of-theart semi-supervised, generative adversarial network (GAN) towards analog performance modeling. We report on various multi-target analog circuit classification experiments and demonstrate stable GAN performance achieving 2-5% higher accuracy and utilizing only 10% fully simulated manually annotated labeled data against supervised learning methods.

IEEE Transactions on Magnetics, 2017

Efficient multihop transmission scheme for error-free relay forwarding in cooperative networks

Wireless Communications and Mobile Computing, 2013

Multihop cooperative communication is emerging as a key concept to extend the coverage area of th... more Multihop cooperative communication is emerging as a key concept to extend the coverage area of the network and potentially increase the capacity. The spectral efficiency of such networks can be improved by adapting the transmission to time-varying channel conditions, referred to as incremental relaying. Although such incremental relaying concepts are progressively being studied, many challenges, such as erroneous transmissions by intermediate nodes and end-to-end delay of the network, limit its practical use due to lack of an efficient implementation. This paper proposes an efficient multihop incremental relaying technique. In this method, erroneous relay forwarding is mitigated, and the overhead for coordination among nodes is reduced by exploiting the implicit feedback channel available due to the broadcast nature of wireless transmissions. The proposed scheme fully leverages the benefit of overhearing and eliminates the additional feedback slots required for validation. Further, it ensures reliable forwarding of information, which optimizes the throughput of multihop networks. Thorough analysis of the proposed scheme is performed under different deployment environments, and the theoretical analyses presented in this paper are supported with results from extensive simulation studies. Copyright

Abstract—Cooperation among nodes is proposed as an effective means of combating fading and for en... more Abstract—Cooperation among nodes is proposed as an effective means of combating fading and for enhancing system’s overall capacity and coverage. The robustness of rateless codes makes them particularly attractive for such networks. Yet, the decoding aspects of rateless codes are discarded in most of the traditional distributed networks. In this paper, the intermediate packet decodability of rateless codes is exploited to reduce the number of packets being processed at the relay nodes. This is achieved by harnessing the back channel (from destination to relays) for feedback. The relay transmissions and processing are thereby confined to assist the receiver in decoding the remaining information. It is shown both analytically and through simulation studies that such a scheme achieves significant savings in computation complexity, memory usage and overall energy consumption. I.

IEEE Transactions on Magnetics

2013 IEEE 24th Annual International Symposium on Personal, Indoor, and Mobile Radio Communications (PIMRC), 2013

ABSTRACT

The Joint Viterbi detector decoder (JVDD) has been proposed as an alternative to the iterative de... more The Joint Viterbi detector decoder (JVDD) has been proposed as an alternative to the iterative detector, performing both detection and decoding in two stages on a trellis. The first stage estimates and retains a set of survivors, while the second stage performs a parity check on these to compute the minimum metric legal codeword (MMLC). With this structure, near optimal maximum-likelihood decoding (MLD) performance can be achieved but at the cost of complexity especially at long codeword lengths (CWL). JVDD codes have been introduced with the explicit target of reducing this complexity. Further, lower rate codes with more parity checks leads to reduced number of survivors in the JVDD trellis resulting in lower complexity. However, it has been observed that JVDD code performance degrades at low-rates while operating in the low SNR region through an error-floor. This aspect is analyzed in this paper and this performance can be attributed to the JVDD code structure where a constant num...

2020 IEEE 31st Magnetic Recording Conference (TMRC), 2020

The hard disk drive (HDD) industry is facing a physical limit on the areal density (AD) of one-di... more The hard disk drive (HDD) industry is facing a physical limit on the areal density (AD) of one-dimensional magnetic recording (1DMR) on traditional magnetic media. To increase capacity without media redesign, twodimensional magnetic recording (TDMR) has been introduced. The effective channel model has a media noise term which models signal dependent noise due to, e.g., magnetic grains intersected by bit boundaries. Trellis based detection with pattern dependent noise prediction (PDNP) [1] is standard practice in HDDs. The trellis detector sends soft coded bit estimates to a channel decoder, which outputs user information bit estimates. PDNP uses a relatively simple autoregressive noise model and linear prediction; this model is somewhat restrictive and may not accurately represent the media noise, especially at high storage densities. To address this modeling problem, we design and train deep neural network (DNN) based media noise predictors. As DNN [2] models are more general than autoregressive models, they more accurately model media noise compared to PDNP. The proposed turbo detector assumes a channel model for the k th linear equalizer filter output y(k):

2017 IEEE 85th Vehicular Technology Conference (VTC Spring), 2017

The rapid growth of commercial cellular and personal communication systems has increased the dema... more The rapid growth of commercial cellular and personal communication systems has increased the demand for high data rate, power and bandwidth efficient transmissions. Advanced coding and modulation techniques are suitable candidates for this purpose as the potential to reduce the symbol period is often limited by the multipath-induced intersymbol interference (ISI). In this paper, a scheme that jointly performs detection, demodulation, and decoding is proposed and is coined the joint detector demodulator decoder (JDDD). The JDDD is a generalization of the previously proposed joint Viterbi detector decoder (JVDD) that performs detection and decoding of binary phase shift keying (BPSK) modulated signals over ISI channels. The JVDD algorithm performs two operations on a trellis: the first operation computes metrics and retains a set of survivors, while the second operation performs parity checking on the survivors to return the minimum metric legal codeword (MMLC). The JDDD algorithm also has a similar structure but the number of states in JDDD trellis is a factor of the number of symbols defined by the modulator and the characteristics of the ISI channel. In this paper, the performance of the proposed JDDD algorithm is evaluated over an ISI channel.

2019 IEEE Intelligent Transportation Systems Conference (ITSC), 2019

IEEE Transactions on Magnetics, 2021

Data detection in magnetic recording (MR) channels can be viewed as an image processing problem, ... more Data detection in magnetic recording (MR) channels can be viewed as an image processing problem, proceeding from the 2-D image of readback bits, to higher level abstractions of features using convolutional layers that finally allow classification of individual bits. In this work, convolutional neural networks (ConvNets) are employed in place of the typical partial response equalizer and maximum-likelihood detector with noise prediction to directly process the un-equalized readback signals and output soft estimates. Several variations of ConvNets are compared in terms of network complexity and performance. The best performing ConvNet detector with two convolutional layers provides a data storage density of up to 3.7489 Terabits/in2 on a low track pitch two-dimensional MR channel simulated with a grain-flipping-probability (GFP) model. An alternate ConvNet architecture reduces the network complexity by about 74%, yet results in only a 2.09% decrease in density compared to the best performing detector.