Reduced overhead error compensation for energy efficient machine learning kernels

Keyphrases

• 45 Nm CMOS Technology 7%
• Algorithmic Noise Tolerance 100%
• Classification Engine 23%
• Classification System 7%
• CMOS Process 7%
• Conventional Architecture 7%
• Dot Product Kernels 15%
• EEG Seizure 7%
• Energy Consumption Reduction 7%
• Energy Reduction 7%
• Energy Saving 7%
• Energy-efficient Machine Learning 100%
• Error Compensation 100%
• Error Rate 7%
• Error Resilience 7%
• Error Tolerance 15%
• Explicit Estimators 7%
• False Positive Rate 7%
• Feature Classification 15%
• Feature Extractor 38%
• Filter Kernel 7%
• Finite Impulse Response Filters 15%
• Learning Kernels 100%
• Low Overhead 15%
• Machine Learning 7%
• Machine Learning Applications 7%
• Magnitude Error 7%
• Order of Magnitude 7%
• Seizure Classification 7%
• Signal Machine 15%
• Signal Processing 15%
• SVM Classification 7%
• True Positive Rate 7%
• Voltage Limit 7%

Engineering

• Dot Product 15%
• Error Compensation 100%
• Error Magnitude 7%
• Error Rate 7%
• Error Resiliency 7%
• Error Tolerance 15%
• Extractor 38%
• Finite Impulse Response Filter 15%
• Learning System 100%
• Noise Tolerance 100%
• Reducing Energy Consumption 7%
• Reformulation 7%
• Simulation Result 7%