Ankur Singh

Ankur Singh, Piyush Rai

Super-Resolution is the technique to improve the quality of a low-resolution photo by boosting its plausible resolution. The computer vision community has extensively explored the area of Super-Resolution. However, previous Super-Resolution methods require vast amounts of data for training which becomes problematic in domains where very few low-resolution, high-resolution pairs might be available. One such area is statistical downscaling, where super-resolution is increasingly being used to obtain high-resolution climate information from low-resolution data. Acquiring high-resolution climate data is extremely expensive and challenging. To reduce the cost of generating high-resolution climate information, Super-Resolution algorithms should be able to train with a limited number of low-resolution, high-resolution pairs. This paper tries to solve the aforementioned problem by introducing a semi-supervised way to perform super-resolution that can generate sharp, high-resolution images with as few as 500 paired examples. The proposed semi-supervised technique can be used as a plug-and-play module with any supervised GAN-based Super-Resolution method to enhance its performance. We quantitatively and qualitatively analyze the performance of the proposed model and compare it with completely supervised methods as well as other unsupervised techniques. Comprehensive evaluations show the superiority of our method over other methods on different metrics. We also offer the applicability of our approach in statistical downscaling to obtain high-resolution climate images.

Ankur Singh, Senthilnath Jayavelu

Despite the recent success of deep neural networks, there remains a need for effective methods to enhance domain generalization using vision transformers. In this paper, we propose a novel domain generalization technique called Robust Representation Learning with Self-Distillation (RRLD) comprising i) intermediate-block self-distillation and ii) augmentation-guided self-distillation to improve the generalization capabilities of transformer-based models on unseen domains. This approach enables the network to learn robust and general features that are invariant to different augmentations and domain shifts while effectively mitigating overfitting to source domains. To evaluate the effectiveness of our proposed method, we perform extensive experiments on PACS and OfficeHome benchmark datasets, as well as an industrial wafer semiconductor defect dataset. The results demonstrate that RRLD achieves robust and accurate generalization performance. We observe an average accuracy improvement in the range of 1.2% to 2.3% over the state-of-the-art on the three datasets.

Subhradip Chakraborty, Ankur Singh, Akhilesh R. Jaiswal

With the rapid growth of deep neural networks (DNNs), compute-in-memory (CIM) has emerged as a promising energy-efficient paradigm for accelerating multiply-and-accumulate (MAC) operations. Yet, current CIM architectures are largely limited to dot-product computations and struggle to efficiently support general-purpose matrix operations, such as transpose, element-wise addition, and multiplication. This work presents a 3D-integrated, memory-on-memory SRAM-eDRAM hybrid CIM architecture, implemented in GlobalFoundries 22~nm FDSOI technology, capable of performing general matrix operations directly within the memory crossbar with 4-bit precision. By leveraging a specialized transpose-based architecture, in-memory arithmetic operations, peripheral-aware design, and 3D SRAM--eDRAM integration, the proposed architecture balances latency, energy efficiency, and compute density for general purpose matrix operations while remaining compatible with the conventional CIM dot product architectures. Overall, this memory-on-memory CIM framework generalizes CIM beyond dot products, enabling versatile matrix processing and paving the way for broader applications in AI acceleration and general-purpose high performance computing.

Ankur Singh, Akhilesh Jaiswal

Read-only memory (ROM) provides deterministic access to predefined data mappings. Extending ROM concepts to the optical domain enables high-bandwidth, low-latency, and parallel memory access, but realizing compact and reconfigurable optical ROM remains challenging due to loss, wavelength control, and integration constraints. This work presents a high-speed, reconfigurable photonic ROM architecture implemented using integrated microring resonators (MRRs). The ROM encodes predefined input-output mappings directly in the spectral response of the photonic devices, enabling deterministic lookup-based operation without dynamic computation during readout. To improve scalability and reduce cumulative insertion loss, the architecture employs compact banked sub-arrays that are selectively addressed through an optical decoding mechanism. Reconfigurability is achieved using transistor-based optical selectors, allowing different ROM banks to be activated without physical light rerouting or interferometric structures. The proposed photonic ROM is designed and evaluated using device-level simulations based on the GlobalFoundries 45SPCLO silicon photonics platform. Simulation results demonstrate reliable operation at data rates up to 12.5 GHz, with stable light-to-current transfer characteristics obtained through integrated photodiode readout. The optical ROM can be used to implement nonlinear activation functions utilised in photonic accelerator architectures, including sigmoid, tanh, ReLU, and exponential mappings.

Ankur Singh, Sanghyeon Choi, Gunuk Wang et al.

Reservoir computing (RC) offers a neuromorphic framework that is particularly effective for processing spatiotemporal signals. Known for its temporal processing prowess, RC significantly lowers training costs compared to conventional recurrent neural networks. A key component in its hardware deployment is the ability to generate dynamic reservoir states. Our research introduces a novel dual-memory RC system, integrating a short-term memory via a WOx-based memristor, capable of achieving 16 distinct states encoded over 4 bits, and a long-term memory component using a TiOx-based memristor within the readout layer. We thoroughly examine both memristor types and leverage the RC system to process temporal data sets. The performance of the proposed RC system is validated through two benchmark tasks: isolated spoken digit recognition with incomplete inputs and Mackey-Glass time series prediction. The system delivered an impressive 98.84% accuracy in digit recognition and sustained a low normalized root mean square error (NRMSE) of 0.036 in the time series prediction task, underscoring its capability. This study illuminates the adeptness of memristor-based RC systems in managing intricate temporal challenges, laying the groundwork for further innovations in neuromorphic computing.

Ankur Singh

The thesis investigates the utilization of memristive and memcapacitive crossbar arrays in low-power machine learning accelerators, offering a comprehensive co-design framework for deep neural networks (DNN). The model, implemented through a hybrid Python and PyTorch approach, accounts for various non-idealities, achieving exceptional training accuracies of 90.02% and 91.03% for the CIFAR-10 dataset with memristive and memcapacitive crossbar arrays on an 8-layer VGG network. Additionally, the thesis introduces a novel approach to emulate meminductor devices using Operational Transconductance Amplifiers (OTA) and capacitors, showcasing adjustable behavior. Transistor-level simulations in 180 nm CMOS technology, operating at 60 MHz, demonstrate the proposed meminductor emulator's viability with a power consumption of 0.337 mW. The design is further validated in neuromorphic circuits and CNN accelerators, achieving training and testing accuracies of 91.04% and 88.82%, respectively. Notably, the exclusive use of MOS transistors ensures the feasibility of monolithic IC fabrication. This research significantly contributes to the exploration of advanced hardware solutions for efficient and high-performance machine-learning applications.

Ankur Singh, Dowon Kim, Byung-Geun Lee

Data-intensive computing tasks, such as training neural networks, are crucial for artificial intelligence applications but often come with high energy demands. One promising solution is to develop specialized hardware that directly maps neural networks, utilizing arrays of memristive devices to perform parallel multiply-accumulate operations. In our research, we introduce a novel CMOS-based memcapacitor circuit that is validated using the cadence tool. Additionally, we developed the device in Python to facilitate the design of a memcapacitive-based accelerator. Our proposed framework employs a crossbar array of memcapacitor devices to train a neural network capable of digit classification and CIFAR dataset recognition. We tested the non-ideal characteristics of the constructed memcapacitor-based neural network. The system achieved an impressive 98.4% training accuracy in digit recognition and 94.4% training accuracy in CIFAR recognition, highlighting its effectiveness. This study demonstrates the potential of memcapacitor-based neural network systems in handling classification tasks and sets the stage for further advancements in neuromorphic computing.

J. Senthilnath, Adithya Bhattiprolu, Ankur Singh et al.

A novel online clustering algorithm is presented where an Evolving Restricted Boltzmann Machine (ERBM) is embedded with a Kohonen Network called ERBM-KNet. The proposed ERBM-KNet efficiently handles streaming data in a single-pass mode using the ERBM, employing a bias-variance strategy for neuron growing and pruning, as well as online clustering based on a cluster update strategy for cluster prediction and cluster center update using KNet. Initially, ERBM evolves its architecture while processing unlabeled image data, effectively disentangling the data distribution in the latent space. Subsequently, the KNet utilizes the feature extracted from ERBM to predict the number of clusters and updates the cluster centers. By overcoming the common challenges associated with clustering algorithms, such as prior initialization of the number of clusters and subpar clustering accuracy, the proposed ERBM-KNet offers significant improvements. Extensive experimental evaluations on four benchmarks and one industry dataset demonstrate the superiority of ERBM-KNet compared to state-of-the-art approaches.

Rajdeep Dutta, Qincheng Wang, Ankur Singh et al.

This paper presents a novel RL algorithm, S-REINFORCE, which is designed to generate interpretable policies for dynamic decision-making tasks. The proposed algorithm leverages two types of function approximators, namely Neural Network (NN) and Symbolic Regressor (SR), to produce numerical and symbolic policies, respectively. The NN component learns to generate a numerical probability distribution over the possible actions using a policy gradient, while the SR component captures the functional form that relates the associated states with the action probabilities. The SR-generated policy expressions are then utilized through importance sampling to improve the rewards received during the learning process. We have tested the proposed S-REINFORCE algorithm on various dynamic decision-making problems with low and high dimensional action spaces, and the results demonstrate its effectiveness and impact in achieving interpretable solutions. By leveraging the strengths of both NN and SR, S-REINFORCE produces policies that are not only well-performing but also easy to interpret, making it an ideal choice for real-world applications where transparency and causality are crucial.

Dwarikanath Mahapatra, Ankur Singh

While medical image segmentation is an important task for computer aided diagnosis, the high expertise requirement for pixelwise manual annotations makes it a challenging and time consuming task. Since conventional data augmentations do not fully represent the underlying distribution of the training set, the trained models have varying performance when tested on images captured from different sources. Most prior work on image synthesis for data augmentation ignore the interleaved geometric relationship between different anatomical labels. We propose improvements over previous GAN-based medical image synthesis methods by learning the relationship between different anatomical labels. We use a weakly supervised segmentation method to obtain pixel level semantic label map of images which is used learn the intrinsic relationship of geometry and shape across semantic labels. Latent space variable sampling results in diverse generated images from a base image and improves robustness. We use the synthetic images from our method to train networks for segmenting COVID-19 infected areas from lung CT images. The proposed method outperforms state-of-the-art segmentation methods on a public dataset. Ablation studies also demonstrate benefits of integrating geometry and diversity.

Sarath Sivaprasad, Ankur Singh, Naresh Manwani et al.

In this paper, we investigate a constrained formulation of neural networks where the output is a convex function of the input. We show that the convexity constraints can be enforced on both fully connected and convolutional layers, making them applicable to most architectures. The convexity constraints include restricting the weights (for all but the first layer) to be non-negative and using a non-decreasing convex activation function. Albeit simple, these constraints have profound implications on the generalization abilities of the network. We draw three valuable insights: (a) Input Output Convex Neural Networks (IOC-NNs) self regularize and reduce the problem of overfitting; (b) Although heavily constrained, they outperform the base multi layer perceptrons and achieve similar performance as compared to base convolutional architectures and (c) IOC-NNs show robustness to noise in train labels. We demonstrate the efficacy of the proposed idea using thorough experiments and ablation studies on standard image classification datasets with three different neural network architectures.

Ankur Singh

Although deep learning performs really well in a wide variety of tasks, it still suffers from catastrophic forgetting -- the tendency of neural networks to forget previously learned information upon learning new tasks where previous data is not available. Earlier methods of incremental learning tackle this problem by either using a part of the old dataset, by generating exemplars or by using memory networks. Although, these methods have shown good results but using exemplars or generating them, increases memory and computation requirements. To solve these problems we propose an adversarial discriminator based method that does not make use of old data at all while training on new tasks. We particularly tackle the class incremental learning problem in image classification, where data is provided in a class-based sequential manner. For this problem, the network is trained using an adversarial loss along with the traditional cross-entropy loss. The cross-entropy loss helps the network progressively learn new classes while the adversarial loss helps in preserving information about the existing classes. Using this approach, we are able to outperform other state-of-the-art methods on CIFAR-100, SVHN, and MNIST datasets.

Ankur Singh, Anurag Chanani, Harish Karnick

In this paper, we tackle the problem of colorization of grayscale videos to reduce bandwidth usage. For this task, we use some colored keyframes as reference images from the colored version of the grayscale video. We propose a model that extracts keyframes from a colored video and trains a Convolutional Neural Network from scratch on these colored frames. Through the extracted keyframes we get a good knowledge of the colors that have been used in the video which helps us in colorizing the grayscale version of the video efficiently. An application of the technique that we propose in this paper, is in saving bandwidth while sending raw colored videos that haven't gone through any compression. A raw colored video takes up around three times more memory size than its grayscale version. We can exploit this fact and send a grayscale video along with out trained model instead of a colored video. Later on, in this paper we show how this technique can help to save bandwidth usage to upto three times while transmitting raw colored videos.