220 likes | 241 Views
Learning Logistic Circuits. Yitao Liang, Guy Van den Broeck. January 31, 2019. Which model to choose. Neural Networks:. Classical AI Methods:. Hungry?. $25?. Sleep?. Restaurant?. …. “Black Box” Good performance on Image Classification. Clear Modeling Assumption. 1. SPN.
E N D
Learning Logistic Circuits Yitao Liang, Guy Van den Broeck January 31, 2019
Which model to choose Neural Networks: Classical AI Methods: Hungry? $25? Sleep? Restaurant? • … • “Black Box” • Good performance on Image Classification • Clear Modeling Assumption 1
SPN Starting Point: Probabilistic Circuits A promising synthesis of the two State-of-the-art on Density Estimation 2
Logical Circuits 1 1 0 Input: 1 1 0 1 0 1 0 0 = 1 AND 0 1 0 1 1 0 0 1 0 Bottom-up Evaluation 0 1 1 0 0 1 4
Logical -> Probabilistic Circuits Red Parameters: Conditional Probabilities 5
Logical -> Probabilistic Circuits Input: 0 0.096 1 0 0.096 0.194 0.24= 0.8*0.3 0.01 0.24 0.00 0.1= 0.1*1 + 0.9*0 0.0 0.3 0.1 0.8 Multiply the parameters bottom-up 0 1 0 1 1 0 0 0 1 0 6
Evaluate Logistic Circuits Input: 0 1 Multiply the parameters bottom-up Logistic function on final output 0 1 1 0 0 1 7
Are logistic circuits amenable to tractable learning 8
Special Case: Logistic Regression Logistic Regression What about other logistic circuits in more general forms? 9
Parameter Learning “Hot” wires are active features 10
Parameter Learning Due to decomposability and determinism, reduce to logistic regression Features associated with each wire “Global Circuit Flow” Convex Parameter learning 11
Structure Learning Similar to LearnPsdd Split nodes to reduce variance of gradients 12
Probabilistic -> Logistic Circuits Probabilities become log-odds Discriminative Counterparts 16
What do Features Mean This is the feature that contributes the most to this image’s classification probability feature value : 0.925 feature weight : 3.489 feature interpretation: curvy lines and hallow center 17
Conclusion Logistic circuits: • Synthesis of symbolic AI and statistical learning • Discriminative counterparts of probabilistic circuits • Convex parameter learning • Simple heuristic for structure learning • Good performance • Easy to interpret 18
Thanks https://github.com/UCLA-StarAI/LogisticCircuit