arXiv subjects

Sayeed Shafayet Chowdhury

Publications and source records attributed to Sayeed Shafayet Chowdhury.

28 records · Page 2Linked to original sources

arXiv2021

One Timestep is All You Need: Training Spiking Neural Networks with Ultra Low Latency

Spiking Neural Networks (SNNs) are energy efficient alternatives to commonly used deep neural networks (DNNs). Through event-driven information processing, SNNs can reduce the expensive compute requirements of DNNs considerably, while achieving comparable performance. However, high inference latency is a significant hindrance to the edge deployment of deep SNNs. Computation over multiple timesteps not only increases latency as well as overall energy budget due to higher number of operations, but also incurs memory access overhead of fetching membrane potentials, both of which lessen the energy benefits of SNNs. To overcome this bottleneck and leverage the full potential of SNNs, we propose an Iterative Initialization and Retraining method for SNNs (IIR-SNN) to perform single shot inference in the temporal axis. The method starts with an SNN trained with T timesteps (T>1). Then at each stage of latency reduction, the network trained at previous stage with higher timestep is utilized as initialization for subsequent training with lower timestep. This acts as a compression method, as the network is gradually shrunk in the temporal domain. In this paper, we use direct input encoding and choose T=5, since as per literature, it is the minimum required latency to achieve satisfactory performance on ImageNet. The proposed scheme allows us to obtain SNNs with up to unit latency, requiring a single forward pass during inference. We achieve top-1 accuracy of 93.05%, 70.15% and 67.71% on CIFAR-10, CIFAR-100 and ImageNet, respectively using VGG16, with just 1 timestep. In addition, IIR-SNNs perform inference with 5-2500X reduced latency compared to other state-of-the-art SNNs, maintaining comparable or even better accuracy. Furthermore, in comparison with standard DNNs, the proposed IIR-SNNs provide25-33X higher energy efficiency, while being comparable to them in classification performance.

cs.NE↗

arXiv2021

Spatio-Temporal Pruning and Quantization for Low-latency Spiking Neural Networks

Spiking Neural Networks (SNNs) are a promising alternative to traditional deep learning methods since they perform event-driven information processing. However, a major drawback of SNNs is high inference latency. The efficiency of SNNs could be enhanced using compression methods such as pruning and quantization. Notably, SNNs, unlike their non-spiking counterparts, consist of a temporal dimension, the compression of which can lead to latency reduction. In this paper, we propose spatial and temporal pruning of SNNs. First, structured spatial pruning is performed by determining the layer-wise significant dimensions using principal component analysis of the average accumulated membrane potential of the neurons. This step leads to 10-14X model compression. Additionally, it enables inference with lower latency and decreases the spike count per inference. To further reduce latency, temporal pruning is performed by gradually reducing the timesteps while training. The networks are trained using surrogate gradient descent based backpropagation and we validate the results on CIFAR10 and CIFAR100, using VGG architectures. The spatiotemporally pruned SNNs achieve 89.04% and 66.4% accuracy on CIFAR10 and CIFAR100, respectively, while performing inference with 3-30X reduced latency compared to state-of-the-art SNNs. Moreover, they require 8-14X lesser compute energy compared to their unpruned standard deep learning counterparts. The energy numbers are obtained by multiplying the number of operations with energy per operation. These SNNs also provide 1-4% higher robustness against Gaussian noise corrupted inputs. Furthermore, we perform weight quantization and find that performance remains reasonably stable up to 5-bit quantization.

cs.LG↗

arXiv2020

DCT-SNN: Using DCT to Distribute Spatial Information over Time for Learning Low-Latency Spiking Neural Networks

Spiking Neural Networks (SNNs) offer a promising alternative to traditional deep learning frameworks, since they provide higher computational efficiency due to event-driven information processing. SNNs distribute the analog values of pixel intensities into binary spikes over time. However, the most widely used input coding schemes, such as Poisson based rate-coding, do not leverage the additional temporal learning capability of SNNs effectively. Moreover, these SNNs suffer from high inference latency which is a major bottleneck to their deployment. To overcome this, we propose a scalable time-based encoding scheme that utilizes the Discrete Cosine Transform (DCT) to reduce the number of timesteps required for inference. DCT decomposes an image into a weighted sum of sinusoidal basis images. At each time step, the Hadamard product of the DCT coefficients and a single frequency base, taken in order, is given to an accumulator that generates spikes upon crossing a threshold. We use the proposed scheme to learn DCT-SNN, a low-latency deep SNN with leaky-integrate-and-fire neurons, trained using surrogate gradient descent based backpropagation. We achieve top-1 accuracy of 89.94%, 68.3% and 52.43% on CIFAR-10, CIFAR-100 and TinyImageNet, respectively using VGG architectures. Notably, DCT-SNN performs inference with 2-14X reduced latency compared to other state-of-the-art SNNs, while achieving comparable accuracy to their standard deep learning counterparts. The dimension of the transform allows us to control the number of timesteps required for inference. Additionally, we can trade-off accuracy with latency in a principled manner by dropping the highest frequency components during inference.

cs.LG↗

arXiv2020

Numerical Analysis of a Highly Sensitive Surface Plasmon Resonance Sensor for SARS-CoV-2 Detection

In this paper, we propose a surface plasmon resonance (SPR) structure based on Kretschmann configuration incorporating layers of Silicon and BaTiO$_3$ on top of Ag for real-time detection of severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) using thiol tethered DNA as a ligand. Extensive numerical analysis based on transfer matrix theory as well as finite-difference time-domain (FDTD) technique has been performed to characterize the sensor response considering sensitivity, full width at half maxima, and minimum reflection. About 7.6 times enhanced sensitivity has been obtained using the proposed architecture for SARS-CoV-2 detection, compared to the basic Kretschmann configuration. Notably, the structure provides consistent enhancement over other competitive SPR structures for both angular and wavelength interrogations with a figure-of-merit of 692.28. Additionally, we repeated simulations for various ligate-ligand pairs to assess the range of applicability and robust performance improvement has been observed. As a result, the proposed sensor design provides a suitable configuration for highly sensitive, rapid, non-invasive biosensing which can be useful if adopted in experimental sensing protocols.

physics.app-ph↗

arXiv2020

Anomaly Detection in Unsupervised Surveillance Setting Using Ensemble of Multimodal Data with Adversarial Defense

Autonomous aerial surveillance using drone feed is an interesting and challenging research domain. To ensure safety from intruders and potential objects posing threats to the zone being protected, it is crucial to be able to distinguish between normal and abnormal states in real-time. Additionally, we also need to consider any device malfunction. However, the inherent uncertainty embedded within the type and level of abnormality makes supervised techniques less suitable since the adversary may present a unique anomaly for intrusion. As a result, an unsupervised method for anomaly detection is preferable taking the unpredictable nature of attacks into account. Again in our case, the autonomous drone provides heterogeneous data streams consisting of images and other analog or digital sensor data, all of which can play a role in anomaly detection if they are ensembled synergistically. To that end, an ensemble detection mechanism is proposed here which estimates the degree of abnormality of analyzing the real-time image and IMU (Inertial Measurement Unit) sensor data in an unsupervised manner. First, we have implemented a Convolutional Neural Network (CNN) regression block, named AngleNet to estimate the angle between a reference image and current test image, which provides us with a measure of the anomaly of the device. Moreover, the IMU data are used in autoencoders to predict abnormality. Finally, the results from these two pipelines are ensembled to estimate the final degree of abnormality. Furthermore, we have applied adversarial attack to test the robustness and security of the proposed approach and integrated defense mechanism. The proposed method performs satisfactorily on the IEEE SP Cup-2020 dataset with an accuracy of 97.8%. Additionally, we have also tested this approach on an in-house dataset to validate its robustness.

eess.IV↗

arXiv2020

Unsupervised Abnormality Detection Using Heterogeneous Autonomous Systems

Anomaly detection (AD) in a surveillance scenario is an emerging and challenging field of research. For autonomous vehicles like drones or cars, it is immensely important to distinguish between normal and abnormal states in real-time. Additionally, we also need to detect any device malfunction. But the nature and degree of abnormality may vary depending upon the actual environment and adversary. As a result, it is impractical to model all cases a-priori and use supervised methods to classify. Also, an autonomous vehicle provides various data types like images and other analog or digital sensor data, all of which can be useful in anomaly detection if leveraged fruitfully. To that effect, in this paper, a heterogeneous system is proposed which estimates the degree of abnormality of an unmanned surveillance drone, analyzing real-time image and IMU (Inertial Measurement Unit) sensor data in an unsupervised manner. Here, we have demonstrated a Convolutional Neural Network (CNN) architecture, named AngleNet to estimate the angle between a normal image and another image under consideration, which provides us with a measure of anomaly of the device. Moreover, the IMU data are used in autoencoder to predict abnormality. Finally, the results from these two algorithms are ensembled to estimate the final degree of abnormality. The proposed method performs satisfactorily on the IEEE SP Cup-2020 dataset with an accuracy of 97.3%. Additionally, we have also tested this approach on an in-house dataset to validate its robustness.

cs.CV↗

arXiv2020

Optimization of Surface Plasmon Resonance Biosensor for Analysis of Lipid Molecules

Surface Plasmon Resonance (SPR) is an important bio-sensing technique for real-time label-free detection. However, it is pivotal to optimize various parameters of the sensor configuration for efficient and highly sensitive sensing. To that effect, we focus on optimizing two different SPR structures -- the basic Kretschmann configuration and narrow groove grating. Our analysis aims to detect two different types of lipids known as phospholipid and eggyolk, which are used as analyte (sensing layer) and two different types of proteins namely tryptophan and bovine serum albumin (BSA) are used as ligand (binding site). For both the configurations, we investigate all possible lipid-protein combinations to understand the effect of various parameters on sensitivity, minimum reflectivity and full width half maximum (FWHM). Lipids are the structural building block of cell membranes and mutation of these layers by virus and bacteria is one the prime reasons of many diseases in our body. Hence, improving the performance of a SPR sensor to detect very small change in lipid holds immense significance. We use finite-difference time-domain (FDTD) technique to perform quantitative analysis to get an optimized structure. We find that sensitivity increases when lipid concentration is increased and it is the highest (21.95 degree/RIU) for phospholipid and tryptophan combination when metal and lipid layer thickness are 45 nm and 30 nm respectively. However, metal layer thickness does not cause any significant variation in sensitivity, but as it increases to 50 nm, minimum reflectivity and full width half maximum (FWHM) decreases to the lowest. In case of narrow groove grating structure, broad range of wavelengths can generate SPR and the sensitivity is highest (900nm/RIU) for a configuration of 10 nm groove width and 70 nm groove height at a resonance wavelength of 1411 nm.

physics.bio-ph↗

arXiv2020

Towards Understanding the Effect of Leak in Spiking Neural Networks

Spiking Neural Networks (SNNs) are being explored to emulate the astounding capabilities of human brain that can learn and compute functions robustly and efficiently with noisy spiking activities. A variety of spiking neuron models have been proposed to resemble biological neuronal functionalities. With varying levels of bio-fidelity, these models often contain a leak path in their internal states, called membrane potentials. While the leaky models have been argued as more bioplausible, a comparative analysis between models with and without leak from a purely computational point of view demands attention. In this paper, we investigate the questions regarding the justification of leak and the pros and cons of using leaky behavior. Our experimental results reveal that leaky neuron model provides improved robustness and better generalization compared to models with no leak. However, leak decreases the sparsity of computation contrary to the common notion. Through a frequency domain analysis, we demonstrate the effect of leak in eliminating the high-frequency components from the input, thus enabling SNNs to be more robust against noisy spike-inputs.

cs.NE↗

arXiv2019

Numerical Analysis of Sensitivity Enhancement of Surface Plasmon Resonance Biosensors Using a Mirrored Bilayer Structure

A mirrored bilayer structure incorporating layers of Graphene and MoS2 is proposed here for Surface Plasmon Resonance (SPR) biosensing and its performance is evaluated numerically. Starting from the basic configuration, the structure with graphene and MoS2 layers is gradually developed for enhanced performance. Reflectance is the main considered parameter for performance analysis. A theoretical framework based on Fresnel's equations is presented and by measuring reflectance versus angle of incidence, sensitivity is calculated from the displacement of SPR angle using finite-difference time-domain (FDTD) technique. Our numerical analysis shows that using the proposed approach, about 4.2 times enhanced sensitivity can be achieved compared to the basic Kretschmann configuration. Notably, the structure provides an enhancement for both angular and wavelength interrogations. Furthermore, simulations have been repeated for various ligate-ligand pairs and consistent enhancement has been observed which proves the robustness of the structure. So, the proposed sensor architecture clearly provides pronounced improvement in sensitivity which can be significant in various practical applications.

physics.optics↗

arXiv2016

The 10% Gd and Ti co-doped BiFeO$_3$: A promising multiferroic material

In this investigation, undoped BiFeO$_3$, Gd doped Bi$_{0.9}$Gd$_{0.1}$FeO$_3$, and Gd-Ti co-doped Bi$_{0.9}$Gd$_{0.1}$Fe$_{1-x}$Ti$_x$O$_3$ (x = 0.10, 0.20) materials were synthesized to report their multiferroic properties. The structural analysis and phase identification of these multiferroic ceramics were performed using Rietveld refinement. The Rietveld analysis has confirmed the high phase purity of the 10% Gd-Ti co-doped Bi$_{0.9}$Gd$_{0.1}$Fe$_{0.9}$Ti$_{0.1}$O$_3$ sample compared to that of other compositions under investigation. The major phase of this particular composition is of rhombohedral \textit{R3c} type structure (wt% $>99\%$) with negligible amount of impurity phases. In terms of characterization, we address magnetic properties of this co-doped ceramic system by applying substantially higher magnetic fields than that applied in previously reported investigations. The dependence of temperature and maximum applied magnetic fields on their magnetization behavior have also been investigated. Additionally, the leakage current density has been measured to explore its effect on the ferroelectric properties of this multiferroic system. The outcome of this investigation suggests that the substitution of 10% Gd and Ti in place of Bi and Fe, respectively, in BiFeO$_3$ significantly enhances its multiferroic properties. The improved properties of this specific composition is associated with homogeneous reduced grain size, significant suppression of impurity phases and reduction in leakage current density which is further asserted by polarization vs. electric field hysteresis loop measurements.

cond-mat.mtrl-sci↗