pPIM: A Programmable Processor-in-Memory Architecture with Precision-Scaling for Deep Learning

Keyphrases

• Deep Learning 100%
• Processing-in-memory Architecture 100%
• Processing-in-memory 100%
• Precision Scaling 100%
• Programmable Processor 100%
• Table-driven 42%
• Logic-based 28%
• Programmability 28%
• Processing Unit 28%
• Convolutional Neural Network 14%
• Low Power 14%
• Energy Efficiency 14%
• Power Consumption 14%
• Low Latency 14%
• Graphics Processing Unit 14%
• Memory Infrastructure 14%
• Memory Write 14%
• Online Training 14%
• Low Quality Data 14%
• Data Access 14%
• Training Implementation 14%
• Data-intensive Applications 14%
• Bandwidth Limitation 14%
• Memory Latency 14%
• Unit Power Consumption 14%
• Logic Circuitry 14%
• Data Transfer 14%
• Processing Speed 14%
• Processor Architecture 14%
• Convolutional Neural Network Application 14%
• Fabrication Challenges 14%
• Ultra-low Power 14%
• Novel Look-up Table (N-LUT) 14%
• Hardware Solution 14%
• Von Neumann Architecture 14%
• Complex Processing 14%
• Precision Operation 14%

Computer Science

• Deep Learning Method 100%
• Memory Architecture 100%
• In-Memory Processing 100%
• Convolutional Neural Network 28%
• Power Consumption 28%
• Processing Unit 28%
• Programmability 28%
• Graphics Processing Unit 14%
• Data Access 14%
• Data Intensive Application 14%
• Energy Efficiency 14%
• Neural Network Application 14%
• Von Neumann Architecture 14%
• Logic Circuitry 14%
• Latency Operation 14%
• Memory Access Latency 14%
• Data Transfer Bandwidth 14%
• Processor Architecture 14%
• Processing Speed 14%