0.25
0.5
0.75
1.25
1.5
1.75
2
Inductive Higher Order Embeddings
Related categories
Presentation
Sometimes Knowledge Graphs are not enough!00:00
Sometimes Knowledge Graphs are not enough!00:34
Sometimes Knowledge Graphs are not enough!01:00
Sometimes Knowledge Graphs are not enough!01:26
Nested Knowledge Graphs01:31
Nested Knowledge Graphs01:45
How do we reason over Nested Knowledge Graphs?02:15
How do we reason over Nested Knowledge Graphs?02:37
How do we reason over Nested Knowledge Graphs?02:59
Sometimes static Nested Knowledge Graphs are not enough!03:19
Sometimes static Nested Knowledge Graphs are not enough!03:58
Sometimes static Nested Knowledge Graphs are not enough!04:10
Sometimes static Nested Knowledge Graphs are not enough!04:17
Sometimes static Nested Knowledge Graphs are not enough!04:21
Inductive Higher Order Embeddings04:30
INKE: Inductive Higher Order Embeddings04:39
Handling Nested Triples: Recursive Flattening05:30
INKE: data model overview – Layer 106:38
INKE: data model overview – Layer 207:27
INKE: data model overview – Layer 308:11
State-of-the-art: KGE approaches08:56
State-of-the-art: KGE approaches09:23
Why multi-layer modeling?09:53
Why multi-layer modeling?10:06
Why multi-layer modeling?10:32
Inductive Reasoning: Handling Unseen Elements10:43
Training: Multi-Level Loss Function11:10
Training: Multi-Level Loss Function11:26
Experimental evaluation: datasets11:42
Baselines11:55
Baselines12:25
Results: Triple Prediction12:30
Results: Triple Prediction12:40
Results: Conditional link prediction13:11
Results: Inductive Reasoning13:36
Conclusions14:07
Limitations and Future Work14:52
Thank you15:34