A5103237172- Advanced Memory and Neural Computing
- Advanced Neural Network Applications
- Numerical methods in engineering
- Parallel Computing and Optimization Techniques
- Composite Material Mechanics
- Ferroelectric and Negative Capacitance Devices
- Machine Learning and ELM
- Ultrasonics and Acoustic Wave Propagation
- Adhesion, Friction, and Surface Interactions
- Fluid dynamics and aerodynamics studies
- Human Pose and Action Recognition
- Advanced Data Storage Technologies
- Mechanical Behavior of Composites
- Advancements in Semiconductor Devices and Circuit Design
- Stochastic Gradient Optimization Techniques
- Video Surveillance and Tracking Methods
- Advancements in Photolithography Techniques
- Fluid Dynamics and Mixing
- Aerodynamics and Fluid Dynamics Research
- Surface Modification and Superhydrophobicity
- Diagnosis and treatment of tuberculosis
- Fault Detection and Control Systems
- Adversarial Robustness in Machine Learning
- Rheology and Fluid Dynamics Studies
- Fatigue and fracture mechanics
Tsinghua University
1996-2024
Recently CNN-based methods have made remarkable progress in broad fields. Both network pruning algorithms and hardware accelerators been introduced to accelerate CNN. However, existing not fully studied the pattern method, current index storage scheme of sparse CNN is efficient. Furthermore, performance suffers from no-load PEs on networks. This work proposes a software-hardware co-design address these problems. The software includes an ADMM-based method which compresses patterns convolution...
This work presents an energy-efficient CIM SoC with heterogeneous CPU, CIM, SIMD, DMA and COMM cores. The main contributions include: 1) A producer-consumer instruction dependency controller (PCIDC) shared multi-port SRAM to reduce SoC-level data transfer. 2) An inner-pipelined read-free digital macro achieve higher frequency. 3) parallel-to-serial (PTS) sparse architecture utilizing the low activity of partial-sum accumulation. demonstrates first digital-CIM SoC. fabricated 55nm chip...
Deploying neural network (NN) models on Internet-of-Things (IoT) devices is important to enable artificial intelligence (AI) the edge realizing AI-of-Things (AIoT). However, high energy consumption and bandwidth requirement of NN restricts AI applications battery-limited equipments. Compute-In-Memory (CIM), featured with efficiency, provides new opportunities for IoT deployment NN. design CIM-based full system still at early stage, lacking system-level demonstration vertical optimization...
In recent years, convolutional neural networks (CNNs) have achieved significant advancements in various fields. However, the computation and storage overheads of CNNs are overwhelming for Internet-of-Things devices. Both network pruning algorithms hardware accelerators been introduced to empower CNN inference at edge. Network reduce size computational cost by regularizing unimportant weights zeros. existing works lack intrakernel structured types tradeoff between sparsity efficiency, index...
Resistive random access memory (RRAM) is an emerging device for processing-in-memory (PIM) architecture to accelerate convolutional neural network (CNN). However, due the highly coupled crossbar structure in RRAM array, it difficult exploit CNN sparsity feature improve performance RRAM-based accelerator. To optimize weight mapping of sparse array and area energy efficiency, we propose a novel scheme corresponding accelerator based on pattern pruning operation unit(OU) mechanism. Experimental...
Resistive Random Access Memory (RRAM) is an emerging device for processing-in-memory (PIM) architecture to accelerate convolutional neural network (CNN). However, due the highly coupled crossbar structure in RRAM array, it difficult exploit sparsity of RRAM-based CNN accelerator. To optimize weight mapping sparse array and achieve high area energy efficiency, we propose a novel scheme corresponding accelerator based on pattern pruning Operation Unit(OU) mechanism. Experimental results show...
Compute-In-Memory (CIM), characterized by efficient matrix-vector multiplication, has been recognized as a promising candidate technology for edge AI computing. However, applying CIM in extreme scenarios, where power delivery is limited and unstable, still faces challenges. The relatively high memory write energy, compared with computing, prevents its further gains on ultra-lower-power devices. frequent backup/restore intermittent together the nonvolatile (NVM) even higher program escalates...
Bionic non-smooth structures on the bodies of revolution had characteristic drag reduction. In this paper, oil flow visualization test were made in low speed wind tunnel between two bionic surface models and smooth model. The results show that pattern is significantly difference according to configuration structures, obviously effect friction However, total coefficient BNNS model reduced. It was found BNSS can decrease pressure obviously. mechanism through sacrifice small viscous force,...
Abstract Background Differentiating between ulcerative colitis (UC), Crohn’s disease (CD) and intestinal tuberculosis (ITB) is challenging under endoscopy. We aimed to realise automatic differential diagnosis among these diseases through machine learning algorithms. Methods A total of 6399 consecutive patients (5128 UC, 875 CD 396 ITB) who had taken colonoscopy examinations in Peking Union Medical College Hospital from January 2008 November 2018 was enrolled. The input the description...