A5078801750- Advanced Neural Network Applications
- Parallel Computing and Optimization Techniques
- Ferroelectric and Negative Capacitance Devices
- Advanced Memory and Neural Computing
- Semantic Web and Ontologies
- Topic Modeling
- Fluid Dynamics Simulations and Interactions
- Data Mining Algorithms and Applications
- Neural Networks and Reservoir Computing
- Advanced Graph Neural Networks
- Graph Theory and Algorithms
- Reservoir Engineering and Simulation Methods
- Natural Language Processing Techniques
- Network Packet Processing and Optimization
- Cell Image Analysis Techniques
- Explainable Artificial Intelligence (XAI)
- Fractal and DNA sequence analysis
- Age of Information Optimization
- Robotic Path Planning Algorithms
- Algorithms and Data Compression
- Advanced Data Storage Technologies
- Neural Networks and Applications
- Stochastic Gradient Optimization Techniques
Harvard University Press
2022-2025
Harvard University
2024
University of Southern California
2022
The field of neuromorphic computing holds great promise in terms advancing efficiency and capabilities by following brain-inspired principles. However, the rich diversity techniques employed research has resulted a lack clear standards for benchmarking, hindering effective evaluation advantages strengths methods compared to traditional deep-learning-based methods. This paper presents collaborative effort, bringing together members from academia industry, define benchmarks computing:...
We introduce QuArch, a dataset of 1500 human-validated question-answer pairs designed to evaluate and enhance language models' understanding computer architecture. The covers areas including processor design, memory systems, performance optimization. Our analysis highlights significant gap: the best closed-source model achieves 84% accuracy, while top small open-source reaches 72%. observe notable struggles in interconnection networks, benchmarking. Fine-tuning with QuArch improves accuracy...
Real-time multi-task multi-model (MTMM) workloads, a new form of deep learning inference are emerging for applications areas like extended reality (XR) to support metaverse use cases. These workloads combine user interactivity with computationally complex machine (ML) activities. Compared standard ML applications, these present unique difficulties and constraints. MTMM impose heterogeneity concurrency requirements on future systems devices, necessitating the development capabilities. This...
Computation on sparse data is becoming increasingly important for many applications. Recent computation accelerators are designed specific algorithm/application, making them inflexible with software optimizations. This paper proposes SparseCore, the first general-purpose processor extension that can flexibly accelerate complex code patterns and fast-evolving algorithms. We extend instruction set architecture (ISA) to make stream or vector first-class citizens, develop efficient architectural...
Graph pattern mining applications try to find all embeddings that match specific patterns. Compared the traditional graph computation, are computation-intensive. The state-of-the-art method, enumeration, constructs pattern. key operation -- intersection of two edge lists, poses challenges conventional architectures and requires substantial execution time. In this paper, we propose IntersectX, a vertically designed accelerator for enumeration with stream instruction set extension...
High-dimensional motion generation requires nu-merical precision for smooth, collision-free solutions. Typically, double-precision or single-precision floating-point (FP) formats are utilized. Using these big tensors imposes a strain on the memory bandwidth provided by devices and alters footprint, hence limiting their applicability to low-power edge needed mobile robots. The uniform application of reduced can be advantageous but severely degrades decreased data types important tensors, we...
Graph Neural Networks (GNNs) are an emerging class of machine learning models which utilize structured graph information and node features to reduce high-dimensional input data low-dimensional embeddings, from predictions can be made. Due the compounding effect aggregating neighbor information, GNN inferences require raw many times more nodes than targeted for prediction. Thus, on heterogeneous compute platforms, inference latency largely subject inter-device communication cost transferring...
High-dimensional motion generation requires numerical precision for smooth, collision-free solutions. Typically, double-precision or single-precision floating-point (FP) formats are utilized. Using these big tensors imposes a strain on the memory bandwidth provided by devices and alters footprint, hence limiting their applicability to low-power edge needed mobile robots. The uniform application of reduced can be advantageous but severely degrades decreased data types important tensors, we...