The TWIML AI Podcast (formerly This Week in Machine Learning & Artificial Intelligence) Deep Learning is Eating 5G. Here’s How, w/ Joseph Soriaga - #525
Oct 7, 2021
Joseph Soriaga, Senior Director of Technology at Qualcomm, explores the exciting intersection of deep learning and 5G technology. He discusses groundbreaking research on augmenting Kalman filters to enhance model efficiency and interpretability. Moreover, he unveils WiCluster, a method for passive indoor positioning using WiFi, shedding light on how AI can optimize 5G networks. Soriaga also highlights the transformative potential of machine learning in delivering connected services, paving the way for a more efficient and interconnected future.
AI Snips
Chapters
Transcript
Episode notes
Shannon Capacity
- Shannon Capacity characterizes the theoretical bandwidth limit between communicating devices.
- It helps determine when to optimize or move on, based on achievable potential.
Neural Augmentation
- Neural augmentation adapts existing algorithms with neural networks to handle mismatches in real-world scenarios.
- This approach leverages domain knowledge while improving prediction accuracy.
Doppler Effect and Kalman Filters
- Doppler effect, caused by movement, impacts channel dynamics in wireless communication.
- Kalman filters, augmented by neural networks, effectively track these dynamic changes.