Link
Embed
Share
Beginning
This slide
Copy link URL
Copy link URL
Copy iframe embed code
Copy iframe embed code
Copy javascript embed code
Copy javascript embed code
Share
Tweet
Share
Tweet
Slide 1
Slide 1 text
Similarity-Based Reconstruction Loss for Meaning Representation
Slide 2
Slide 2 text
Literature 2
Slide 3
Slide 3 text
Abstract • • • 3
Slide 4
Slide 4 text
Introduction • • 4
Slide 5
Slide 5 text
Related Work • • • • 5
Slide 6
Slide 6 text
Related Work • • 6
Slide 7
Slide 7 text
Auto-Encoder •ℒ , • • • • 7
Slide 8
Slide 8 text
Weighted similarity loss •ℒ = − σ =1 sim , • • • : • • sim() • 8
Slide 9
Slide 9 text
Weighted cross-entropy loss •ℒ = − σ =1 sim , log( ) • • 9
Slide 10
Slide 10 text
Soft label loss •ℒ = − σ =1 ∗log • ∗ = ൞ sim , σ =1 sim(,) , ∈ top N 0 , ∉ top N • • 10
Slide 11
Slide 11 text
True-label encoding 11
Slide 12
Slide 12 text
Tasks & Datasets • • • 12
Slide 13
Slide 13 text
Results 13
Slide 14
Slide 14 text
Results 14
Slide 15
Slide 15 text
Additional Experiments • • 15
Slide 16
Slide 16 text
Results • • 16
Slide 17
Slide 17 text
Results 17
Slide 18
Slide 18 text
Results 18
Slide 19
Slide 19 text
Results 19
Slide 20
Slide 20 text
Discussion • • 20
Slide 21
Slide 21 text
Conclusion • • • • 21