XMA: A Crossbar-aware Multi-task Adaption Framework via Shift-based Mask Learning Method

Keyphrases

• Backbone Model 20%
• Binary Mask 20%
• Catastrophic Forgetting 20%
• Convolution Engine 20%
• Crossbar 100%
• Crossbar Array 20%
• Deep Neural Network Accelerator 20%
• Deep Neural Network Inference 20%
• Design Innovation 20%
• Effective Structure 20%
• Energy Efficient 20%
• High Energy Consumption 20%
• High Energy Efficiency 20%
• Learning Algorithm 20%
• Learning Methods 100%
• Mask Based 20%
• Mask Learning 100%
• Masking Scheme 20%
• Memory Overhead 40%
• Minimal Hardware 20%
• Multi-task 100%
• Need for Power 20%
• New Task 40%
• Piggyback 40%
• Popular 20%
• ReRAM Cells 20%
• ReRAM Crossbar 20%
• ReRAM Crossbar Array 20%
• Specific Task 20%
• Task-oriented 20%

Computer Science

• Computer Hardware 33%
• Convolution 33%
• Deep Neural Network 66%
• Energy Consumption 33%
• Energy Efficiency 33%
• Energy Efficient 33%
• Experimental Result 33%
• Learning Algorithm 33%
• Multiple Task 100%
• Network Inference 33%