On-device next-item recommendation with self-supervised knowledge distillation

Keyphrases

• Recommender Systems 100%
• Next-item Recommendation 100%
• On-device 100%
• Knowledge Distillation 100%
• Tensor Decomposition 66%
• Public Concern 33%
• Popular 33%
• Compress 33%
• Original Model 33%
• Capacity Loss 33%
• Upload 33%
• Fundamental Challenges 33%
• User Profile 33%
• Decomposition Method 33%
• User Needs 33%
• Size Reduction 33%
• User Interest 33%
• Behavioral Data 33%
• Carbon Footprint 33%
• Environmental Context 33%
• Recommendation Model 33%
• Ultra-compact 33%
• Long-term Users 33%
• Higher-order Process 33%
• Performance Degradation 33%
• Session-based Recommender Systems 33%
• Long Tail Items 33%
• GitHub 33%
• Item Recommendation 33%
• Memory Resource 33%
• Computing Resources 33%
• Regular Model 33%
• User Request 33%
• Accuracy Loss 33%
• Recombination Strategies 33%
• Compression Rate 33%
• Compact Model 33%
• HTTPS 33%
• Limited Memory 33%
• Edge Devices 33%
• Concern for Information Privacy 33%
• Concurrent Users 33%
• Local Inference 33%
• Cost-conscious 33%
• User Support 33%

Engineering

• Dimensionality 100%
• Public Concern 100%
• Raw Data 100%
• User Profile 100%
• Performance Degradation 100%
• Carbon Footprint 100%
• Size Reduction 100%

Computer Science

• Recommender Systems 100%
• Knowledge Distillation 100%
• Processing Speed 25%
• Behavior Data 25%
• Compression Rate 25%
• Computing Resource 25%
• Performance Degradation 25%
• Memory Resource 25%
• Environmental Context 25%
• Concurrent User 25%