Yiwei Zhao
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Research Subjects
Yiwei Zhao has a total of 18 co-authors. Here are the top co-authors by number of shared publications:
- Phillip B. Gibbons
Carnegie Mellon University, Pittsburgh, USA
7 shared publications - Hongbo Kang
Tsinghua University, Beijing, China
5 shared publications - Charles McGuffey
Reed College, Portland, USA
4 shared publications - Guy E. Blelloch
Carnegie Mellon University
4 shared publications - Laxman Dhulipala
University of Maryland, College Park, College PARK, USA
4 shared publications - Yan Gu
University of California, Riverside, Riverside, USA
4 shared publications
Latest Publications
- Open Access
- 6 July 2026
- Published by Association for Computing Machinery (ACM)
- p. 392-404
Non-Clairvoyant Scheduling for Processing-in-Memory
- 6 July 2026
- Published by Association for Computing Machinery (ACM)
- p. 392-404
Open AccessNon-Clairvoyant Scheduling for Processing-in-Memory
- Open Access
- 28 January 2026
- Published by Association for Computing Machinery (ACM)
- p. 480-495
PIM-zd-tree: A Fast Space-Partitioning Index Leveraging Processing-in-Memory
- 28 January 2026
- Published by Association for Computing Machinery (ACM)
- p. 480-495
Open AccessPIM-zd-tree: A Fast Space-Partitioning Index Leveraging Processing-in-Memory
- Open Access
- 28 July 2025
- Published by Association for Computing Machinery (ACM)
- p. 28-30
H4H: Hybrid Convolution-Transformer Architecture Search for NPU-CIM Heterogeneous Systems for AR/VR Applications (Abstract)
- 28 July 2025
- Published by Association for Computing Machinery (ACM)
- p. 28-30
Open AccessH4H: Hybrid Convolution-Transformer Architecture Search for NPU-CIM Heterogeneous Systems for AR/VR Applications (Abstract)
- Open Access
- 16 July 2025
- Published by Association for Computing Machinery (ACM)
- p. 350-366
Optimal Batch-Dynamic kd-trees for Processing-in-Memory with Applications
- 16 July 2025
- Published by Association for Computing Machinery (ACM)
- p. 350-366
Open AccessOptimal Batch-Dynamic kd-trees for Processing-in-Memory with Applications
- Open Access
- 20 January 2025
- Published by Association for Computing Machinery (ACM)
- p. 1133-1141
H4H: Hybrid Convolution-Transformer Architecture Search for NPU-CIM Heterogeneous Systems for AR/VR Applications
Low-latency and low-power edge AI is crucial for Augmented/Virtual Reality applications. Recent advances demonstrate that hybrid models, combining convolution layers (CNN) and transformers (ViT), often achieve a superior accuracy/performance tradeoff on various computer vision and machine learning (ML) tasks. However, hybrid ML models can present system challenges for latency and energy efficiency due to their diverse nature in dataflow and memory access patterns. In this work, we leverage architecture heterogeneity from Neural Processing Units (NPU) and Compute-In-Memory (CIM) and explore diverse execution schemas for efficient hybrid model executions. We introduce H4H-NAS, a two-stage Neural Architecture Search (NAS) framework to automate the design of hybrid CNN/ViT models for heterogeneous edge systems featuring both NPU and CIM. We propose a two-phase incremental supernet training in our NAS to resolve gradient conflicts between sampled subnets caused by different block types in a hybrid model search space. Our H4H-NAS approach is also powered by a performance estimator built with NPU performance results measured on real silicon, and CIM performance based on industry IPs. H4H-NAS searches hybrid CNN-ViT models with fine granularity and achieves significant (up to 1.34%) top-1 accuracy improvement on ImageNet-1k. Moreover, results from our algorithm/hardware co-design reveal up to 56.08% overall latency and 41.72% energy improvements by introducing heterogeneous computing over baseline solutions. Overall, our framework guides the design of hybrid network architectures and system architectures for NPU+CIM heterogeneous systems.
- 20 January 2025
- Published by Association for Computing Machinery (ACM)
- p. 1133-1141
Open AccessH4H: Hybrid Convolution-Transformer Architecture Search for NPU-CIM Heterogeneous Systems for AR/VR Applications
Low-latency and low-power edge AI is crucial for Augmented/Virtual Reality applications. Recent advances demonstrate that hybrid models, combining convolution layers (CNN) and transformers (ViT), often achieve a superior accuracy/performance tradeoff on various computer vision and machine learning (ML) tasks. However, hybrid ML models can present system challenges for latency and energy efficiency due to their diverse nature in dataflow and memory access patterns. In this work, we leverage architecture heterogeneity from Neural Processing Units (NPU) and Compute-In-Memory (CIM) and explore diverse execution schemas for efficient hybrid model executions. We introduce H4H-NAS, a two-stage Neural Architecture Search (NAS) framework to automate the design of hybrid CNN/ViT models for heterogeneous edge systems featuring both NPU and CIM. We propose a two-phase incremental supernet training in our NAS to resolve gradient conflicts between sampled subnets caused by different block types in a hybrid model search space. Our H4H-NAS approach is also powered by a performance estimator built with NPU performance results measured on real silicon, and CIM performance based on industry IPs. H4H-NAS searches hybrid CNN-ViT models with fine granularity and achieves significant (up to 1.34%) top-1 accuracy improvement on ImageNet-1k. Moreover, results from our algorithm/hardware co-design reveal up to 56.08% overall latency and 41.72% energy improvements by introducing heterogeneous computing over baseline solutions. Overall, our framework guides the design of hybrid network architectures and system architectures for NPU+CIM heterogeneous systems.
- Open Access
- 18 July 2023
- Published by Association for Computing Machinery (ACM)
- p. 13-14
PIM-tree: A Skew-resistant Index for Processing-in-Memory (Abstract) | Proceedings of the 2023 ACM Workshop on Highlights of Parallel Computing
Processing-in-memory (PIM) is an emerging technology to alleviate the high cost of data movement by pushing computation into/near memory modules. There is an inherent tension, however, between minimizing communication (data movement) and achieving load balance in PIM systems in the presence of workload skew. This work introduces PIM-tree, a PIM-based index that simultaneously achieves low communication, good load balance, and low space consumption. It achieves good theoretical bounds in the PIM Model and efficient on a real-world PIM machine, outperforming prior PIM-based and state-of-the-art CPU-based indexes.
- 18 July 2023
- Published by Association for Computing Machinery (ACM)
- p. 13-14
Open AccessPIM-tree: A Skew-resistant Index for Processing-in-Memory (Abstract) | Proceedings of the 2023 ACM Workshop on Highlights of Parallel Computing
Processing-in-memory (PIM) is an emerging technology to alleviate the high cost of data movement by pushing computation into/near memory modules. There is an inherent tension, however, between minimizing communication (data movement) and achieving load balance in PIM systems in the presence of workload skew. This work introduces PIM-tree, a PIM-based index that simultaneously achieves low communication, good load balance, and low space consumption. It achieves good theoretical bounds in the PIM Model and efficient on a real-world PIM machine, outperforming prior PIM-based and state-of-the-art CPU-based indexes.
- Open Access
- 17 June 2023
- Published by Association for Computing Machinery (ACM)
- p. 1-14
PIM-trie: A Skew-resistant Trie for Processing-in-Memory
Memory latency and bandwidth are significant bottlenecks in designing in-memory indexes. Processing-in-memory (PIM), an emerging hardware design approach, alleviates this problem by embedding processors in memory modules, enabling low-latency memory access whose aggregated bandwidth scales linearly with the number of PIM modules. Despite recent work in balanced comparison-based indexes on PIM systems, building efficient tries for PIMs remains an open challenge due to tries' inherently unbalanced shape. This paper presents the PIM-trie, the first batch-parallel radix-based index for PIM systems that provides load balance and low communication under adversary-controlled workloads. We introduce trie matching-matching a query trie of a batch against the compressed data trie-as a key building block for PIM-friendly index operations. Our algorithm combines (i) hash-based comparisons for coarse-grained work distribution/elimination and (ii) bit-by-bit comparisons for fine-grained matching. Combined with other techniques (meta-block decomposition, selective recursive replication, differentiated verification), PIM-trie supports LongestCommonPrefix, Insert, and Delete in O(logP) communication rounds per batch and O(l/w) communication volume per string, where P is the number of PIM modules, l is the string length in bits, and w is the machine word size. Moreover, work and communication are load-balanced among modules whp, even under worst-case skew.
- 17 June 2023
- Published by Association for Computing Machinery (ACM)
- p. 1-14
Open AccessPIM-trie: A Skew-resistant Trie for Processing-in-Memory
Memory latency and bandwidth are significant bottlenecks in designing in-memory indexes. Processing-in-memory (PIM), an emerging hardware design approach, alleviates this problem by embedding processors in memory modules, enabling low-latency memory access whose aggregated bandwidth scales linearly with the number of PIM modules. Despite recent work in balanced comparison-based indexes on PIM systems, building efficient tries for PIMs remains an open challenge due to tries' inherently unbalanced shape. This paper presents the PIM-trie, the first batch-parallel radix-based index for PIM systems that provides load balance and low communication under adversary-controlled workloads. We introduce trie matching-matching a query trie of a batch against the compressed data trie-as a key building block for PIM-friendly index operations. Our algorithm combines (i) hash-based comparisons for coarse-grained work distribution/elimination and (ii) bit-by-bit comparisons for fine-grained matching. Combined with other techniques (meta-block decomposition, selective recursive replication, differentiated verification), PIM-trie supports LongestCommonPrefix, Insert, and Delete in O(logP) communication rounds per batch and O(l/w) communication volume per string, where P is the number of PIM modules, l is the string length in bits, and w is the machine word size. Moreover, work and communication are load-balanced among modules whp, even under worst-case skew.