Lecture 20: Scaling Inference with RDBMS Slides by Niu et al.
Today Markov Logic Scaling up grounding with RBDMS Scaling up search with partitioning Main-memory Statistical Analytics
Section 1 1. Markov Logic
Section 1 Data Model
Section 1 Markov Logic*
Markov Logic by Example Section 1 Markov Logic by Example
Section 1 Inference
Section 1 Inference
How to Perform Inference Section 1 How to Perform Inference
How to Perform Inference Section 1 How to Perform Inference
How to Perform Inference Section 1 How to Perform Inference
2. Scaling up Grounding with RDBMS
Section 2 Challenge
Section 2 Grounding in Alchemy*
Section 2 Grounding with RDBMS
Grounding Performance Section 2 Grounding Performance
3. Scaling Search
Section 3 Scaling Search
Section 3 Scaling Search
Partition to Scale up Search Section 3 Partition to Scale up Search
Effect of Partitioning Section 3 Effect of Partitioning
Partitioning vs Quality Section 3 Partitioning vs Quality
Partitioning Improves Quality Section 3 Partitioning Improves Quality
It’s all about tradeoffs Section 3 It’s all about tradeoffs
4. Main-Memory Statistical Analytics
Section 3 Memory Model and SGD
Section 3 Access Methods
Section 3 Modern Architectures
It’s all about tradeoffs Section 3 It’s all about tradeoffs
It’s all about tradeoffs Section 3 It’s all about tradeoffs
It’s all about tradeoffs Section 3 It’s all about tradeoffs
Conclusion Optimized data access makes a difference It’s critical to understand and exploit tradeoffs Modern hardware introduces new opportunities