publications

2023

1. ICLR SNN

   SalientGrads: Sparse Models for Communication Efficient and data aware Distributed Federated Training [coming soon]

   Riyasat Ohib, Bishal Thapaliya, Pratyush Reddy, Jingyu Liu, Vince Calhoun, and Sergey Plis

   ICLR Sparse Neural Networks Workshop 2023

   Abs HTML

   Federated learning (FL) enables training of a model leveraging decentralized data in client sites while preserving privacy by not collecting data. However, one of the significant challenges of FL is limited computation and low communication bandwidth in resource limited edge client nodes. To address this several solutions have been proposed in recent times including transmitting sparse models and learning dynamic masks iteratively among others. However, many of these methods rely on transmitting the model weights throughout the training process as they are based and ad-hoc or random pruning criteria. In this work, we propose \technique which simplifies the process of sparse training by choosing a subnetwork before training based on aggregated model connection saliency scores calculated from the local client data and only sharing highly sparse gradients between server and clients during the training phase. We also demonstrate the efficacy of our method in a real world federated learning framework and report improvement in wall-clock communication time.

2022

1. TMLR

   Explicit Group Sparse Projection with Applications to Deep Learning and NMF

   Riyasat Ohib, Nicolas Gillis, Niccolò Dalmasso, Sameena Shah, Vamsi Potluru, and Sergey Plis

   Transactions on Machine Learning Research 2022

   Abs HTML

   We design a new sparse projection method for a set of vectors that guarantees a desired average sparsity level measured leveraging the popular Hoyer measure (an affine function of the ratio of the \ell_1 and \ell_2 norms). Existing approaches either project each vector individually or require the use of a regularization parameter which implicitly maps to the average \ell_0-measure of sparsity. Instead, in our approach we set the sparsity level for the whole set explicitly and simultaneously project a group of vectors with the sparsity level of each vector tuned automatically. We show that the computational complexity of our projection operator is linear in the size of the problem. Additionally, we propose a generalization of this projection by replacing the \ell_1 norm by its weighted version. We showcase the efficacy of our approach in both supervised and unsupervised learning tasks on image datasets including CIFAR10 and ImageNet. In deep neural network pruning, the sparse models produced by our method on ResNet50 have significantly higher accuracies at corresponding sparsity values compared to existing competitors. In nonnegative matrix factorization, our approach yields competitive reconstruction errors against state-of-the-art algorithms.

2021

1. NeurIPS Off-RL

   Single-Shot Pruning for Offline Reinforcement Learning

   Samin Yeasar, Riyasat Ohib, Sergey Plis, and Doina Precup

   NeurIPS Offline RL Workshop 2021

   Abs HTML PDF

   Deep Reinforcement Learning (RL) is a powerful framework for solving complex real-world problems. Large neural networks employed in the framework are traditionally associated with better generalization capabilities, but their increased size entails the drawbacks of extensive training duration, substantial hardware resources, and longer inference times. One way to tackle this problem is to prune neural networks leaving only the necessary parameters. State-of-the-art concurrent pruning techniques for imposing sparsity perform demonstrably well in applications where data-distributions are fixed. However, they have not yet been substantially explored in the context of RL. We close the gap between RL and single-shot pruning techniques and present a general pruning approach to the Offline RL. We leverage a fixed dataset to prune neural networks before the start of RL training. We then run experiments varying the network sparsity level and evaluating the validity of pruning at initialization techniques in continuous control tasks. Our results show that with 95% of the network weights pruned, Offline-RL algorithms can still retain performance in the majority of our experiments. To the best of our knowledge no prior work utilizing pruning in RL retained performance at such high levels of sparsity. Moreover, pruning at initialization techniques can be easily integrated into any existing Offline-RL algorithms without changing the learning objective.

2. ICLR HAET

   Grouped Sparse Projection for Deep Learning

   Riyasat Ohib, Nicolas Gillis, Sergey Plis, and Vamsi Potluru

   ICLR Hardware Aware Efficient Training workshop 2021

   Abs HTML PDF

   Accumulating empirical evidence shows that very large deep learning models learn faster and achieve higher accuracy than their smaller counterparts. Yet, smaller models have benefits of energy efficiency and are often easier to interpret. To simultaneously get the benefits of large and small models we often encourage sparsity in the model weights of large models. For this, different approaches have been proposed including weight-pruning and distillation. Unfortunately, most existing approaches do not have a controllable way to request a desired value of sparsity as an interpretable parameter and get it right in a single run. In this work, we design a new sparse projection method for a set of weights in order to achieve a desired average level of sparsity without additional hyperparameter tuning which is measured using the ratio of the l1 and l2 norms. Instead of projecting each vector of the weight matrix individually, or using sparsity as a regularizer, we project all vectors together to achieve an average target sparsity, where the sparsity levels of the individual vectors of the weight matrix are automatically tuned. Our projection operator has the following guarantees – (A) it is fast and enjoys a runtime linear in the size of the vectors; (B) the solution is unique except for a measure set of zero. We utilize our projection operator to obtain the desired sparsity of deep learning models in a single run with a negligible performance hit, while competing methods require sparsity hyperparameter tuning. Even with a single projection of a pre-trained dense model followed by fine-tuning, we show empirical performance competitive to the state of the art. We support these claims with empirical evidence on real-world datasets and on a number of architectures, comparing it to other state of the art methods including DeepHoyer.

2017

1. IEEE-Xplore

   Power file extraction process from bangladesh grid and exploring ENF based classification accuracy using machine learning

   Samin Yeasar Arnob, Riyasat Ohib, Md Muhtady Muhaisin, and Tanzil Bin Hassan

   In 2017 IEEE Region 10 Humanitarian Technology Conference (R10-HTC) 2017

   Abs HTML

   The Electric Network Frequency (ENF) is the supply frequency of power distribution networks, which can be captured by multimedia signals recorded near electrical activities. It normally fluctuates slightly over time from its nominal value of 50 Hz/60 Hz. The ENF remain consistent across the entire power grid. This has led to the emergence of multiple forensic application like estimating the recording location and validating the time of recording. Recently an ENF based Machine Learning system was proposed which infers that the region of recording can be identified using ENF signal extracted from the recorded multimedia signal, with the help of relevant features. As supervised learning process requires ground truth to train classifier for identifying future unknown data, in this work- we report Power Recording data extraction process from the National Grid of Bangladesh. Furthermore, we used ENF data – derived from Power Recordings, to compare grids around the world and found out classification accuracy of Bangladesh National Grid. ENF derivation process from Power Recording data and set of features, which serve as identifying characteristics for detecting the region of origin of the multimedia recordingare followed from published work. We used those characteristics in a multiclass Machine Learning implementation based on MATLAB which is able to identify the grid of the recorded signal.

2016

1. IEEE-Xplore

   Metal nanoparticle enhanced light absorption in GaAs thin-film solar cell

   Riyasat Ohib, Samin Yeasar Arnob, Md Sayem Ali, Rakibul Hasan Sagor, and Md Ruhul Amin

   In 2016 IEEE Asia-Pacific Conference on Applied Electromagnetics (APACE) 2016

   Abs HTML

   Surface plasmon resonances in metallic nanoparticles are of considerable importance for thin film technologies due to the significant electromagnetic enhancement in the vicinity of metal surface, light trapping and the ability to tune the resonant wavelength by varying the size, geometry and local dielectric environment of the metal nanoparticle. Metal nanoparticles have the ability for enhanced light trapping, increased absorption and overall rise in efficiency of solar cells. Light enhancement engineered on the vicinity of such particles can augment the absorption in the absorber layer of the Photovoltaic (PV) cell. In this report, we present calculations of light absorption inside the GaAs absorber layer with embedded metal nanospheres and the subsequent effects instigated by the plasmonic phenomena. The finite element simulation software COMSOL Multiphysics was employed for calculations of light absorption near Au, Ag, Cu and Al nanoparticles and optimization of particle diameters. The results conclude that above 600 nm wavelength where the solar radiation is absorbed poorly by GaAs, the absorption with the presence of nanoparticles in GaAs absorber layer can be enhanced by a factor of 2.

2. IEEE-SigPort

   ENF Based Grid Classification System: Identifying the Region of Origin of Digital Recordings

   Riyasat Ohib, Samin Yeasar Arnob, Riazul Arefin, Md Amin, and Taslim Reza

   SigPort: IEEE SP Cup 2016 2016

   Abs HTML

   The Electric Network Frequency (ENF) is the supply frequency of power distribution networks which can be captured by multimedia signals recorded near electrical activities. It normally fluctuates slightly over time from its nominal value, which is usually of 50 Hz/60 Hz. The ENF remain consistent across the entire power grid. This has led to the emergence of multiple forensic application like estimating the recording location and validating the time of recording. In this report we examine an ENF based Machine Learning system which infers the power grid in which the multimedia signal was recorded. We worked on different features which serve as signature for power grid. Then we used those in a multiclass machine learning implementation which is able to identify the grid of the recorded signal.