Vijaykrishnan Narayanan
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Vijaykrishnan Narayanan has a total of 367 co-authors. Here are the top co-authors by number of shared publications:
- Xueqing Li
Tsinghua University
26 shared publications - Suman Datta
Georgia Institute of Technology
23 shared publications - Kai Ni
University of Notre Dame, Notre Dame, IN, USA
20 shared publications John Sampson
Department of Electrical Engineering and Computer Science (EECS), The Pennsylvania State University, University Park, PA, USA
15 shared publications- Yongpan Liu
Tsinghua University
14 shared publications - Huazhong Yang
Tsinghua University
13 shared publications
Latest Publications
- Open Access
- 4 August 2026
- Published by Association for Computing Machinery (ACM)
- p. 23-30
Sigma-Delta Neural Network Conversion on Loihi 2
Neuromorphic computing aims to improve the efficiency of artificial neural networks by taking inspiration from biological neurons and leveraging temporal sparsity, spatial sparsity, and compute near/in memory. Although these approaches have shown efficiency gains, training these spiking neural networks (SNN) remains difficult. The original attempts at converting trained conventional analog neural networks (ANN) to SNNs used the rate of binary spikes to represent neuron activations. This required many simulation time steps per inference, which degraded efficiency. Intel’s Loihi 2 is a neuromorphic platform that supports graded spikes which can be used to represent changes in neuron activation. In this work, we use Loihi 2’s graded spikes to develop a method for converting ANN networks to spiking networks, which exploits temporal and spatial sparsity. We evaluated the performance of this network on Loihi 2 and compared it to NVIDIA’s Jetson Xavier edge AI platform. The results show that neuromorphic approaches achieve significant improvements in efficiency and latency (energy-delay product) over existing solutions.
- 4 August 2026
- Published by Association for Computing Machinery (ACM)
- p. 23-30
Open AccessSigma-Delta Neural Network Conversion on Loihi 2
Neuromorphic computing aims to improve the efficiency of artificial neural networks by taking inspiration from biological neurons and leveraging temporal sparsity, spatial sparsity, and compute near/in memory. Although these approaches have shown efficiency gains, training these spiking neural networks (SNN) remains difficult. The original attempts at converting trained conventional analog neural networks (ANN) to SNNs used the rate of binary spikes to represent neuron activations. This required many simulation time steps per inference, which degraded efficiency. Intel’s Loihi 2 is a neuromorphic platform that supports graded spikes which can be used to represent changes in neuron activation. In this work, we use Loihi 2’s graded spikes to develop a method for converting ANN networks to spiking networks, which exploits temporal and spatial sparsity. We evaluated the performance of this network on Loihi 2 and compared it to NVIDIA’s Jetson Xavier edge AI platform. The results show that neuromorphic approaches achieve significant improvements in efficiency and latency (energy-delay product) over existing solutions.
- Open Access
- 1 June 2026
- Published by IOP Publishing in Neuromorphic Computing and Engineering
- Vol. 6 (2) , 022003
A review of event-based vision sensor fusion: architectures, evaluation and robustness challenges
Event-based vision is a neuromorphic sensing approach that enables low-latency, energy-aware perception by encoding brightness changes as sparse, asynchronous events rather than dense image frames. Event cameras provide microsecond-level timestamps and high dynamic range, supporting perception during fast motion and challenging illumination. In autonomy, event streams are often fused with complementary sensors such as inertial measurement units, color cameras, light-detection-and-ranging sensors, or radar to improve scale estimation, robustness, and redundancy. Event-centric fusion is challenging because sensor clocks and noise models differ across modalities, event rates are activity-dependent, event representations trade timing fidelity for computational convenience, and hardware cost is often dominated by communication and memory movement. Robustness to missing or degraded modalities is common in deployment but remains inconsistently evaluated. This topical review synthesizes event-based vision fusion around three coupled design decisions: fusion stage, temporal coupling, and event representation. We show how these choices determine whether systems preserve event-driven efficiency or revert to dense processing. As a case study, we review event-based depth estimation, consolidate benchmark results, and analyze missing-modality behavior. The evidence suggests that many gains are reported primarily as accuracy improvements, with limited analysis of modality dropout, event-rate shifts, timing mismatch, preprocessing cost, and efficiency. We relate recurring patterns to optical flow and semantic perception, connect them to neuromorphic hardware and hardware-algorithm co-design, and conclude with benchmarking recommendations and open problems in robustness, scalability, and training of spiking and hybrid models.
- 1 June 2026
- Published by IOP Publishing in Neuromorphic Computing and Engineering
- Vol. 6 (2) , 022003
Open AccessA review of event-based vision sensor fusion: architectures, evaluation and robustness challenges
Event-based vision is a neuromorphic sensing approach that enables low-latency, energy-aware perception by encoding brightness changes as sparse, asynchronous events rather than dense image frames. Event cameras provide microsecond-level timestamps and high dynamic range, supporting perception during fast motion and challenging illumination. In autonomy, event streams are often fused with complementary sensors such as inertial measurement units, color cameras, light-detection-and-ranging sensors, or radar to improve scale estimation, robustness, and redundancy. Event-centric fusion is challenging because sensor clocks and noise models differ across modalities, event rates are activity-dependent, event representations trade timing fidelity for computational convenience, and hardware cost is often dominated by communication and memory movement. Robustness to missing or degraded modalities is common in deployment but remains inconsistently evaluated. This topical review synthesizes event-based vision fusion around three coupled design decisions: fusion stage, temporal coupling, and event representation. We show how these choices determine whether systems preserve event-driven efficiency or revert to dense processing. As a case study, we review event-based depth estimation, consolidate benchmark results, and analyze missing-modality behavior. The evidence suggests that many gains are reported primarily as accuracy improvements, with limited analysis of modality dropout, event-rate shifts, timing mismatch, preprocessing cost, and efficiency. We relate recurring patterns to optical flow and semantic perception, connect them to neuromorphic hardware and hardware-algorithm co-design, and conclude with benchmarking recommendations and open problems in robustness, scalability, and training of spiking and hybrid models.
- 5 December 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Electron Device Letters
- Vol. 47 (2) , 395-398
Erase-Verify Operation in Ferroelectric 3D Vertical NAND Storage
Ferroelectric 3D vertical NAND offers attractive benefits such as lower stack height and reduced write voltage, but enabling erase-verify remains a significant challenge because the erase operation places all FeFETs into a high-threshold-voltage (HVT) state, suppressing electron conduction through the string. This makes conventional electron-based erase-verify schemes ineffective. In this work, we propose a GIDL-assisted hole-based erase-verify method that generates holes through band-to-band tunneling in the source-side transistor and uses their propagation to distinguish fully erased (all-HVT) strings from those containing fail-to-erase low-Vth (LVT) cells. TCAD simulations show that GIDL-generated holes can propagate across an all-HVT string, raising the bitline potential, while an LVT FeFET blocks hole transport, enabling robust verify sensing. We further analyze device-level reliability considerations and evaluate the impact of channel thickness on sense margin. The proposed GIDL-assisted approach provides a compact and architecture-compatible solution for erase-verify in FE-NAND without modifying channel type, adding p+ regions, or requiring incremental erase schemes.
- 5 December 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Electron Device Letters
- Vol. 47 (2) , 395-398
Erase-Verify Operation in Ferroelectric 3D Vertical NAND Storage
Ferroelectric 3D vertical NAND offers attractive benefits such as lower stack height and reduced write voltage, but enabling erase-verify remains a significant challenge because the erase operation places all FeFETs into a high-threshold-voltage (HVT) state, suppressing electron conduction through the string. This makes conventional electron-based erase-verify schemes ineffective. In this work, we propose a GIDL-assisted hole-based erase-verify method that generates holes through band-to-band tunneling in the source-side transistor and uses their propagation to distinguish fully erased (all-HVT) strings from those containing fail-to-erase low-Vth (LVT) cells. TCAD simulations show that GIDL-generated holes can propagate across an all-HVT string, raising the bitline potential, while an LVT FeFET blocks hole transport, enabling robust verify sensing. We further analyze device-level reliability considerations and evaluate the impact of channel thickness on sense margin. The proposed GIDL-assisted approach provides a compact and architecture-compatible solution for erase-verify in FE-NAND without modifying channel type, adding p+ regions, or requiring incremental erase schemes.
- Open Access
MoRA: Missing Modality Low-Rank Adaptation for Visual Recognition
Pre-trained vision language models have shown remarkable performance on visual recognition tasks, but they typically assume the availability of complete multimodal inputs during both training and inference. In real-world scenarios, however, modalities may be missing due to privacy constraints, collection difficulties, or resource limitations. While previous approaches have addressed this challenge using prompt learning techniques, they fail to capture the cross-modal relationships necessary for effective multimodal visual recognition and suffer from inevitable computational overhead. In this paper, we introduce MoRA, a parameter-efficient fine-tuning method that explicitly models cross-modal interactions while maintaining modality-specific adaptations. MoRA introduces modality-common parameters between text and vision encoders, enabling bidirectional knowledge transfer. Additionally, combined with the modality-specific parameters, MoRA allows the backbone model to maintain inter-modality interaction and enable intra-modality flexibility. Extensive experiments on standard benchmarks demonstrate that MoRA achieves an average performance improvement in missing-modality scenarios by 5.24% and uses only 25.90% of the inference time compared to the SOTA method while requiring only 0.11% of trainable parameters compared to full fine-tuning.
Open AccessMoRA: Missing Modality Low-Rank Adaptation for Visual Recognition
Pre-trained vision language models have shown remarkable performance on visual recognition tasks, but they typically assume the availability of complete multimodal inputs during both training and inference. In real-world scenarios, however, modalities may be missing due to privacy constraints, collection difficulties, or resource limitations. While previous approaches have addressed this challenge using prompt learning techniques, they fail to capture the cross-modal relationships necessary for effective multimodal visual recognition and suffer from inevitable computational overhead. In this paper, we introduce MoRA, a parameter-efficient fine-tuning method that explicitly models cross-modal interactions while maintaining modality-specific adaptations. MoRA introduces modality-common parameters between text and vision encoders, enabling bidirectional knowledge transfer. Additionally, combined with the modality-specific parameters, MoRA allows the backbone model to maintain inter-modality interaction and enable intra-modality flexibility. Extensive experiments on standard benchmarks demonstrate that MoRA achieves an average performance improvement in missing-modality scenarios by 5.24% and uses only 25.90% of the inference time compared to the SOTA method while requiring only 0.11% of trainable parameters compared to full fine-tuning.
- Open Access
Windsock is Dancing: Adaptive Multimodal Retrieval-Augmented Generation
Multimodal Retrieval-Augmented Generation (MRAG) has emerged as a promising method to generate factual and up-to-date responses of Multimodal Large Language Models (MLLMs) by incorporating non-parametric knowledge from external knowledge bases. However, existing MRAG approaches suffer from static retrieval strategies, inflexible modality selection, and suboptimal utilization of retrieved information, leading to three critical challenges: determining when to retrieve, what modality to incorporate, and how to utilize retrieved information effectively. To address these challenges, we introduce Windsock, a query-dependent module making decisions on retrieval necessity and modality selection, effectively reducing computational overhead and improving response quality. Additionally, we propose Dynamic Noise-Resistance (DANCE) Instruction Tuning, an adaptive training strategy that enhances MLLMs' ability to utilize retrieved information while maintaining robustness against noise. Moreover, we adopt a self-assessment approach leveraging knowledge within MLLMs to convert question-answering datasets to MRAG training datasets. Extensive experiments demonstrate that our proposed method significantly improves the generation quality by 17.07% while reducing 8.95% retrieval times.
Open AccessWindsock is Dancing: Adaptive Multimodal Retrieval-Augmented Generation
Multimodal Retrieval-Augmented Generation (MRAG) has emerged as a promising method to generate factual and up-to-date responses of Multimodal Large Language Models (MLLMs) by incorporating non-parametric knowledge from external knowledge bases. However, existing MRAG approaches suffer from static retrieval strategies, inflexible modality selection, and suboptimal utilization of retrieved information, leading to three critical challenges: determining when to retrieve, what modality to incorporate, and how to utilize retrieved information effectively. To address these challenges, we introduce Windsock, a query-dependent module making decisions on retrieval necessity and modality selection, effectively reducing computational overhead and improving response quality. Additionally, we propose Dynamic Noise-Resistance (DANCE) Instruction Tuning, an adaptive training strategy that enhances MLLMs' ability to utilize retrieved information while maintaining robustness against noise. Moreover, we adopt a self-assessment approach leveraging knowledge within MLLMs to convert question-answering datasets to MRAG training datasets. Extensive experiments demonstrate that our proposed method significantly improves the generation quality by 17.07% while reducing 8.95% retrieval times.
- Open Access
- 22 October 2025
- Published by Association for Computing Machinery (ACM)
- p. 1-9
NaviSense: A Multimodal Assistive Mobile application for Object Retrieval by Persons with Visual Impairment
- 22 October 2025
- Published by Association for Computing Machinery (ACM)
- p. 1-9
Open AccessNaviSense: A Multimodal Assistive Mobile application for Object Retrieval by Persons with Visual Impairment
- Open Access
- 17 October 2025
- Published by Association for Computing Machinery (ACM)
- p. 1178-1192
Multi-Dimensional ML-Pipeline Optimization in Cost-Effective Disaggregated Datacenter
- 17 October 2025
- Published by Association for Computing Machinery (ACM)
- p. 1178-1192
Open AccessMulti-Dimensional ML-Pipeline Optimization in Cost-Effective Disaggregated Datacenter
- Open Access
- 10 October 2025
- Published by Association for Computing Machinery (ACM) in ACM Transactions on Embedded Computing Systems
- Vol. 24 (6) , 1-20
Ferroelectric Memory Technology for Big Data Applications
Big Data has an insatiable appetite for larger and better-performing memory. While current memory technologies continue to advance, the performance gaps in current memory and storage technology have motivated the exploration of emerging memory technologies capable of providing new functionalities. Ferroelectric memory is one such promising candidate which has recently experienced a revival after the discovery of ferroelectricity in hafnium dioxide (HfO 2 )—the dielectric of choice in advanced CMOS manufacturing. While the commercial viability of ferroelectric memory technology has made significant progress over the past decade, several challenges related to variation and reliability still stand as a barrier to large-scale commercial implementation. Here, we review some of the outstanding challenges of ferroelectric memory technology along with the recent materials and device innovations that are being considered to overcome them. Moreover, we aim to highlight these challenges as materials and device co-design problems that must be addressed through collaborative efforts that straddle the two disciplines. We identify and provide our perspective on some of the key challenges and opportunities for ferroelectric-based microelectronic technology.
- 10 October 2025
- Published by Association for Computing Machinery (ACM) in ACM Transactions on Embedded Computing Systems
- Vol. 24 (6) , 1-20
Open AccessFerroelectric Memory Technology for Big Data Applications
Big Data has an insatiable appetite for larger and better-performing memory. While current memory technologies continue to advance, the performance gaps in current memory and storage technology have motivated the exploration of emerging memory technologies capable of providing new functionalities. Ferroelectric memory is one such promising candidate which has recently experienced a revival after the discovery of ferroelectricity in hafnium dioxide (HfO 2 )—the dielectric of choice in advanced CMOS manufacturing. While the commercial viability of ferroelectric memory technology has made significant progress over the past decade, several challenges related to variation and reliability still stand as a barrier to large-scale commercial implementation. Here, we review some of the outstanding challenges of ferroelectric memory technology along with the recent materials and device innovations that are being considered to overcome them. Moreover, we aim to highlight these challenges as materials and device co-design problems that must be addressed through collaborative efforts that straddle the two disciplines. We identify and provide our perspective on some of the key challenges and opportunities for ferroelectric-based microelectronic technology.
- Open Access
- 9 October 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Journal on Exploratory Solid-State Computational Devices and Circuits
- Vol. 11, 148-156
Understanding Reliability Trade-Offs in 1T-nC and 2T-nC FeRAM Designs
Ferroelectric random-access memory (FeRAM) is a promising candidate for energy-efficient non-volatile memory, particularly for logic-in-memory and compute-in-memory applications. Among the available cell architectures, 1T–nC and 2T–nC FeRAMs each offer distinct trade-offs in density, scalability, and reliability. In this work, we present a comparative study of these two architectures under both dimensional scaling (XY/Z shrinkage) and vertical integration (increasing stacked capacitors per cell). Using TCAD and circuit-level simulations, we analyze how scaling impacts ferroelectric capacitance, parasitic coupling, and floating-node dynamics, which together dictate sense margin and read stability. A key mitigation strategy—floating unselected capacitors—is applied to both architectures, effectively decoupling the sense margin from the number of stacked capacitors and enabling tractable analysis across scaling regimes. Results show that 1T–nC suffers more from charge sharing with the bitline, while 2T–nC benefits from transistor isolation and stronger low-voltage sensing at the cost of increased area. By systematically evaluating these behaviors across scaling directions, this work establishes the reliability trade-offs of 1T–nC and 2T–nC cells and provides design guidelines for high-density, vertically integrated FeRAM systems.
- 9 October 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Journal on Exploratory Solid-State Computational Devices and Circuits
- Vol. 11, 148-156
Open AccessUnderstanding Reliability Trade-Offs in 1T-nC and 2T-nC FeRAM Designs
Ferroelectric random-access memory (FeRAM) is a promising candidate for energy-efficient non-volatile memory, particularly for logic-in-memory and compute-in-memory applications. Among the available cell architectures, 1T–nC and 2T–nC FeRAMs each offer distinct trade-offs in density, scalability, and reliability. In this work, we present a comparative study of these two architectures under both dimensional scaling (XY/Z shrinkage) and vertical integration (increasing stacked capacitors per cell). Using TCAD and circuit-level simulations, we analyze how scaling impacts ferroelectric capacitance, parasitic coupling, and floating-node dynamics, which together dictate sense margin and read stability. A key mitigation strategy—floating unselected capacitors—is applied to both architectures, effectively decoupling the sense margin from the number of stacked capacitors and enabling tractable analysis across scaling regimes. Results show that 1T–nC suffers more from charge sharing with the bitline, while 2T–nC benefits from transistor isolation and stronger low-voltage sensing at the cost of increased area. By systematically evaluating these behaviors across scaling directions, this work establishes the reliability trade-offs of 1T–nC and 2T–nC cells and provides design guidelines for high-density, vertically integrated FeRAM systems.
- Open Access
- 6 October 2025
- Published by Association for Computing Machinery (ACM)
- p. 8-14
Revisiting Pebble Games for Modeling and Efficient Use of Disaggregated Memory Systems
- 6 October 2025
- Published by Association for Computing Machinery (ACM)
- p. 8-14
Open AccessRevisiting Pebble Games for Modeling and Efficient Use of Disaggregated Memory Systems