Xiaoxuan Yang
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Xiaoxuan Yang has a total of 79 co-authors. Here are the top co-authors by number of shared publications:
- Yiran Chen
Duke University, Durham, North Carolina, USA
5 shared publications - Hai Li
Duke University, Durham, NC, USA
4 shared publications - Janardhan Rao Doppa
Washington State University
3 shared publications - Krishnendu Chakrabarty
Arizona State University
3 shared publications - Partha Pratim Pande
Washington State University
3 shared publications - Peilin Chen
University of Virginia, Charlottesville, VA, USA
3 shared publications
Latest Publications
- Open Access
- 7 October 2026
- Published by Wiley in Advanced Intelligent Systems
- p. e70573
Deep Learning Classification of Pancreatic Cancer Cells Over a Progression of Epithelial to Mesenchymal States by Deformability Cytometry Under Microfluidic Viscoelastic Flows
Cancer cell classification over intermediate or hybrid states during the epithelial‐to‐mesenchymal transition (EMT) provides information on their heterogeneity, plasticity, and invasiveness. While accomplished by immunofluorescence imaging of protein markers, it is low‐throughput and limited by sample heterogeneity. Since intracellular redistribution of filamentous proteins during EMT alters cellular biomechanics, we present single‐cell imaging over a continuum of deformation and recovery regions under microfluidic viscoelastic flows coupled to a multi‐region deep learning framework for high‐throughput EMT classification on morphometric shape descriptors, binary masks of cell geometry, and brightfield images retaining intracellular texture. We infer that despite nuclear enlargement during EMT that enhances stiffness, vimentin redistribution around the cytoskeleton likely allows progressively EMT‐induced pancreatic cancer cells to support greater deformation and faster relaxation to isotropic shapes. Using an architecture combining convolutional feature extraction with attention‐based regional aggregation, the contributions of cell deformation and relaxation toward classification of intermediate EMT states are captured using global geometric information and intracellular texture related to cytoskeletal remodeling. In comparison to single‐region imaging flow cytometry or scalar deformability metrics, the reported multi‐region attention‐based measurement of the deformation and relaxation dynamics improves the classification of intermediate EMT states that present the greatest plasticity for metastasis.
- 7 October 2026
- Published by Wiley in Advanced Intelligent Systems
- p. e70573
Open AccessDeep Learning Classification of Pancreatic Cancer Cells Over a Progression of Epithelial to Mesenchymal States by Deformability Cytometry Under Microfluidic Viscoelastic Flows
Cancer cell classification over intermediate or hybrid states during the epithelial‐to‐mesenchymal transition (EMT) provides information on their heterogeneity, plasticity, and invasiveness. While accomplished by immunofluorescence imaging of protein markers, it is low‐throughput and limited by sample heterogeneity. Since intracellular redistribution of filamentous proteins during EMT alters cellular biomechanics, we present single‐cell imaging over a continuum of deformation and recovery regions under microfluidic viscoelastic flows coupled to a multi‐region deep learning framework for high‐throughput EMT classification on morphometric shape descriptors, binary masks of cell geometry, and brightfield images retaining intracellular texture. We infer that despite nuclear enlargement during EMT that enhances stiffness, vimentin redistribution around the cytoskeleton likely allows progressively EMT‐induced pancreatic cancer cells to support greater deformation and faster relaxation to isotropic shapes. Using an architecture combining convolutional feature extraction with attention‐based regional aggregation, the contributions of cell deformation and relaxation toward classification of intermediate EMT states are captured using global geometric information and intracellular texture related to cytoskeletal remodeling. In comparison to single‐region imaging flow cytometry or scalar deformability metrics, the reported multi‐region attention‐based measurement of the deformation and relaxation dynamics improves the classification of intermediate EMT states that present the greatest plasticity for metastasis.
- Open Access
- 5 August 2026
- Published by Association for Computing Machinery (ACM)
- p. 1-7
SpikON: A Dual-Parallel and Efficient Accelerator for Online Spiking Neural Networks Learning
Spiking neural networks (SNNs) have emerged as a promising paradigm for energy-efficient brain-inspired computing. However, existing online unsupervised SNN learning suffers from low training accuracy and poor scalability. Although current online supervised learning algorithms perform well on large-scale datasets and networks, the non-hardware-friendly operations hinder efficient edge deployment. In this work, we propose SpikON, the first algorithm-hardware co-design framework for efficient and scalable end-to-end online supervised SNN learning. We first propose the learnable threshold through time and scaled weight centralization through time techniques to address the inefficiency of traditional algorithms. Moreover, to reduce latency and energy consumption, we introduce the novel training dataflow and cascade computation reuse scheme for SNNs that allows concurrent forward-backward computation and temporal reuse across timesteps. We further design the dedicated SNN accelerator with a dual-parallel engine and customized SIMD-based SNN core for efficient end-to-end online learning. Experiments show that the SpikON algorithm achieves 32.2% and 35.0% reductions in training latency and energy consumption over the baseline, without sacrificing accuracy. Moreover, the SpikON co-design achieves 7.2× (11.5×) and 26.8× (15.8×) training throughput (energy efficiency) compared with the edge Apple M4 GPU and TPU-like accelerator, respectively. The code is available at GitHub.
- 5 August 2026
- Published by Association for Computing Machinery (ACM)
- p. 1-7
Open AccessSpikON: A Dual-Parallel and Efficient Accelerator for Online Spiking Neural Networks Learning
Spiking neural networks (SNNs) have emerged as a promising paradigm for energy-efficient brain-inspired computing. However, existing online unsupervised SNN learning suffers from low training accuracy and poor scalability. Although current online supervised learning algorithms perform well on large-scale datasets and networks, the non-hardware-friendly operations hinder efficient edge deployment. In this work, we propose SpikON, the first algorithm-hardware co-design framework for efficient and scalable end-to-end online supervised SNN learning. We first propose the learnable threshold through time and scaled weight centralization through time techniques to address the inefficiency of traditional algorithms. Moreover, to reduce latency and energy consumption, we introduce the novel training dataflow and cascade computation reuse scheme for SNNs that allows concurrent forward-backward computation and temporal reuse across timesteps. We further design the dedicated SNN accelerator with a dual-parallel engine and customized SIMD-based SNN core for efficient end-to-end online learning. Experiments show that the SpikON algorithm achieves 32.2% and 35.0% reductions in training latency and energy consumption over the baseline, without sacrificing accuracy. Moreover, the SpikON co-design achieves 7.2× (11.5×) and 26.8× (15.8×) training throughput (energy efficiency) compared with the edge Apple M4 GPU and TPU-like accelerator, respectively. The code is available at GitHub.
- Open Access
- 22 June 2026
- Published by Association for Computing Machinery (ACM)
- p. 479-486
SpiKint: Native Full-Integer Spiking Neural Networks Training with an Efficient CIM-based Accelerator
- 22 June 2026
- Published by Association for Computing Machinery (ACM)
- p. 479-486
Open AccessSpiKint: Native Full-Integer Spiking Neural Networks Training with an Efficient CIM-based Accelerator
- Open Access
- 29 June 2025
- Published by Association for Computing Machinery (ACM)
- p. 71-77
Titanus: Enabling KV Cache Pruning and Quantization On-the-Fly for LLM Acceleration
- 29 June 2025
- Published by Association for Computing Machinery (ACM)
- p. 71-77
Open AccessTitanus: Enabling KV Cache Pruning and Quantization On-the-Fly for LLM Acceleration
- Open Access
- 29 June 2025
- Published by Association for Computing Machinery (ACM)
- p. 791-797
AutoRAC: Automated Processing-in-Memory Accelerator Design for Recommender Systems
- 29 June 2025
- Published by Association for Computing Machinery (ACM)
- p. 791-797
Open AccessAutoRAC: Automated Processing-in-Memory Accelerator Design for Recommender Systems
- Open Access
- 5 November 2024
- Published by American Chemical Society (ACS) in ACS Electrochemistry
- Vol. 1 (1) , 115-122
Sub-2 nm Amorphous PdSn/C Nanocatalyst for Enhanced Ethanol Oxidation
- 5 November 2024
- Published by American Chemical Society (ACS) in ACS Electrochemistry
- Vol. 1 (1) , 115-122
Open AccessSub-2 nm Amorphous PdSn/C Nanocatalyst for Enhanced Ethanol Oxidation
- 3 June 2024
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Computer-Aided Design of Integrated Circuits and Systems
- Vol. 43 (12) , 4558-4571
Block-Wise Mixed-Precision Quantization: Enabling High Efficiency for Practical ReRAM-Based DNN Accelerators
Resistive random access memory (ReRAM)-based processing-in-memory (PIM) architectures have demonstrated great potential to accelerate Deep Neural Network (DNN) training/ inference. However, the computational accuracy of analog PIM is compromised due to the non-idealities, such as the conductance variation of ReRAM cells. The impact of these non-idealities worsens as the number of concurrently activated wordlines and bitlines increases. To guarantee computational accuracy, only a limited number of wordlines and bitlines of the crossbar array can be turned on concurrently, significantly reducing the achievable parallelism of the architecture. While the constraints on parallelism limit the efficiency of the accelerators, they also provide a new opportunity for finegrained mixed-precision quantization. To enable efficient DNN inference on practical ReRAM-based accelerators, we propose an algorithm-architecture co-design framework called Block-Wise mixed-precision Quantization (BWQ). At the algorithm level, BWQ-A introduces a mixed-precision quantization scheme at the block level, which achieves a high weight and activation compression ratio with negligible accuracy degradation. We also present the hardware architecture design BWQ-H, which leverages the low-bit-width models achieved by BWQ-A to perform high-efficiency DNN inference on ReRAM devices. BWQ-H also adopts a novel precision-aware weight mapping method to increase the ReRAM crossbars throughput. Our evaluation demonstrates the effectiveness of BWQ, which achieves a 6.08× speedup and a 17.47× energy saving on average compared to existing ReRAM-based architectures.
- 3 June 2024
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Computer-Aided Design of Integrated Circuits and Systems
- Vol. 43 (12) , 4558-4571
Block-Wise Mixed-Precision Quantization: Enabling High Efficiency for Practical ReRAM-Based DNN Accelerators
Resistive random access memory (ReRAM)-based processing-in-memory (PIM) architectures have demonstrated great potential to accelerate Deep Neural Network (DNN) training/ inference. However, the computational accuracy of analog PIM is compromised due to the non-idealities, such as the conductance variation of ReRAM cells. The impact of these non-idealities worsens as the number of concurrently activated wordlines and bitlines increases. To guarantee computational accuracy, only a limited number of wordlines and bitlines of the crossbar array can be turned on concurrently, significantly reducing the achievable parallelism of the architecture. While the constraints on parallelism limit the efficiency of the accelerators, they also provide a new opportunity for finegrained mixed-precision quantization. To enable efficient DNN inference on practical ReRAM-based accelerators, we propose an algorithm-architecture co-design framework called Block-Wise mixed-precision Quantization (BWQ). At the algorithm level, BWQ-A introduces a mixed-precision quantization scheme at the block level, which achieves a high weight and activation compression ratio with negligible accuracy degradation. We also present the hardware architecture design BWQ-H, which leverages the low-bit-width models achieved by BWQ-A to perform high-efficiency DNN inference on ReRAM devices. BWQ-H also adopts a novel precision-aware weight mapping method to increase the ReRAM crossbars throughput. Our evaluation demonstrates the effectiveness of BWQ, which achieves a 6.08× speedup and a 17.47× energy saving on average compared to existing ReRAM-based architectures.
- 23 November 2023
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Circuits and Systems II: Express Briefs
- Vol. 71 (3) , 1683-1689
Neuro-Symbolic Computing: Advancements and Challenges in Hardware–Software Co-Design
The rapid progress of artificial intelligence (AI) has led to the emergence of a highly promising field known as neuro-symbolic (NeSy) computing. This approach combines the strengths of neural networks, which excel at data-driven learning, with the reasoning capabilities of symbolic AI. Neuro-symbolic models have the potential to overcome the limitations of each approach individually, resulting in interpretable and explainable AI systems that can reason over complex knowledge bases, learn from limited and/or noisy data, and be generalizable. However, the exploration of NeSy AI from a system perspective remains limited. This paper provides an in-depth analysis of the state-ofthe-art hardware-software co-design techniques for NeSy AI and discusses the associated challenges in improving system efficiency for heterogeneous computing. By examining the intersection of NeSy computing and system design, we aim to bridge the gap and foster advancements in this domain.
- 23 November 2023
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Circuits and Systems II: Express Briefs
- Vol. 71 (3) , 1683-1689
Neuro-Symbolic Computing: Advancements and Challenges in Hardware–Software Co-Design
The rapid progress of artificial intelligence (AI) has led to the emergence of a highly promising field known as neuro-symbolic (NeSy) computing. This approach combines the strengths of neural networks, which excel at data-driven learning, with the reasoning capabilities of symbolic AI. Neuro-symbolic models have the potential to overcome the limitations of each approach individually, resulting in interpretable and explainable AI systems that can reason over complex knowledge bases, learn from limited and/or noisy data, and be generalizable. However, the exploration of NeSy AI from a system perspective remains limited. This paper provides an in-depth analysis of the state-ofthe-art hardware-software co-design techniques for NeSy AI and discusses the associated challenges in improving system efficiency for heterogeneous computing. By examining the intersection of NeSy computing and system design, we aim to bridge the gap and foster advancements in this domain.
- Open Access
Block-Wise Mixed-Precision Quantization: Enabling High Efficiency for Practical ReRAM-based DNN Accelerators
Resistive random access memory (ReRAM)-based processing-in-memory (PIM) architectures have demonstrated great potential to accelerate Deep Neural Network (DNN) training/inference. However, the computational accuracy of analog PIM is compromised due to the non-idealities, such as the conductance variation of ReRAM cells. The impact of these non-idealities worsens as the number of concurrently activated wordlines and bitlines increases. To guarantee computational accuracy, only a limited number of wordlines and bitlines of the crossbar array can be turned on concurrently, significantly reducing the achievable parallelism of the architecture. While the constraints on parallelism limit the efficiency of the accelerators, they also provide a new opportunity for fine-grained mixed-precision quantization. To enable efficient DNN inference on practical ReRAM-based accelerators, we propose an algorithm-architecture co-design framework called \underline{B}lock-\underline{W}ise mixed-precision \underline{Q}uantization (BWQ). At the algorithm level, BWQ-A introduces a mixed-precision quantization scheme at the block level, which achieves a high weight and activation compression ratio with negligible accuracy degradation. We also present the hardware architecture design BWQ-H, which leverages the low-bit-width models achieved by BWQ-A to perform high-efficiency DNN inference on ReRAM devices. BWQ-H also adopts a novel precision-aware weight mapping method to increase the ReRAM crossbar's throughput. Our evaluation demonstrates the effectiveness of BWQ, which achieves a 6.08x speedup and a 17.47x energy saving on average compared to existing ReRAM-based architectures.
Open AccessBlock-Wise Mixed-Precision Quantization: Enabling High Efficiency for Practical ReRAM-based DNN Accelerators
Resistive random access memory (ReRAM)-based processing-in-memory (PIM) architectures have demonstrated great potential to accelerate Deep Neural Network (DNN) training/inference. However, the computational accuracy of analog PIM is compromised due to the non-idealities, such as the conductance variation of ReRAM cells. The impact of these non-idealities worsens as the number of concurrently activated wordlines and bitlines increases. To guarantee computational accuracy, only a limited number of wordlines and bitlines of the crossbar array can be turned on concurrently, significantly reducing the achievable parallelism of the architecture. While the constraints on parallelism limit the efficiency of the accelerators, they also provide a new opportunity for fine-grained mixed-precision quantization. To enable efficient DNN inference on practical ReRAM-based accelerators, we propose an algorithm-architecture co-design framework called \underline{B}lock-\underline{W}ise mixed-precision \underline{Q}uantization (BWQ). At the algorithm level, BWQ-A introduces a mixed-precision quantization scheme at the block level, which achieves a high weight and activation compression ratio with negligible accuracy degradation. We also present the hardware architecture design BWQ-H, which leverages the low-bit-width models achieved by BWQ-A to perform high-efficiency DNN inference on ReRAM devices. BWQ-H also adopts a novel precision-aware weight mapping method to increase the ReRAM crossbar's throughput. Our evaluation demonstrates the effectiveness of BWQ, which achieves a 6.08x speedup and a 17.47x energy saving on average compared to existing ReRAM-based architectures.
- 20 January 2023
- Published by Elsevier in Carbohydrate Polymers
- Vol. 307, 120600
High-stretchable, self-healing, self-adhesive, self-extinguishing, low-temperature tolerant starch-based gel and its application in stimuli-responsiveness
- 20 January 2023
- Published by Elsevier in Carbohydrate Polymers
- Vol. 307, 120600
High-stretchable, self-healing, self-adhesive, self-extinguishing, low-temperature tolerant starch-based gel and its application in stimuli-responsiveness