A Unified Memory Perspective for Probabilistic Trustworthy AI

1. 1Understand the challenges of probabilistic computation in AI systems
2. 2Identify performance bottlenecks in memory systems
3. 3Develop a unified memory perspective to address these bottlenecks
4. 4Apply this perspective to improve the robustness and efficiency of AI models

AI engineers and researchers benefit from this work as it aims to improve the robustness and efficiency of probabilistic AI systems, while data scientists and ML researchers can apply these concepts to their models

💡 Probabilistic AI workloads shift performance bottlenecks from arithmetic units to memory systems

💡 Unified memory perspective for probabilistic trustworthy AI