-
### GAT link prediction on unseen nodes
*Hi! I am trying to implement link prediction based on GAT and predict unseen nodes (Inductive learning). While I can successfully run a link prediction usin…
-
### Description
We have two sorts of knowledge graph algorithms:
1. ComplEx, DistMult, ...: using `KGTripleGenerator`, which passes in node and edge type ilocs to look up in embedding layers
2.…
huonw updated
3 years ago
-
In GAE, the decoder is trying to reconstruct the adjacency matrix. I presume that this approach makes the GVAE completely transductive because an unseen node will not be present in the adjacency matr…
-
The GraphSage implements structural embeddings, but requires calculation of the eigenvalues and eigenvectors to do so. This is time and memory intensive, and requires (re-)calculation of eigenvalues a…
-
Note: the issue was created automatically with bugzilla2github tool
Original bug ID: BZ#2779
From: @braibant
Reported version: 8.5
CC: @aspiwack, cauger@lri.fr, @pirbo, @braibant
-
#### Version
8.8 master hash 8dd6d09 and 8.7.0
#### Description of the problem
```
Inductive foo (a : bool) (a : nat) : Type
:= bar : @foo ltac:(assumption) ltac:(assumption).
Print foo.…
-
## 一言でいうと
構造が異なるモデル間で蒸留を行うことで、特定モデルで学習しやすい知識(CNNなら局所特徴、RNNなら系列構造など)を転移できるか検証した研究。CNN=>MLP、LSTM=>Transformerで蒸留を行いそれぞれの学習傾向が蒸留先モデルに反映されることを確認。
### 論文リンク
https://arxiv.org/abs/2006.00555
###…
-
### Prerequisites
Please put an X between the brackets as you perform the following steps:
* [X] Check that your issue is not already filed:
https://github.com/leanprover/lean4/issues
* …
-
we need a libobject which populates a map kername -> loc (where loc contains a file name relative to the logical path)
mathcomp.ssreflect.ssrnat.len -> mathcomp.ssreflect + ssrnat.v + loc
Coq.Init…
gares updated
1 month ago
-
たとえば Nat.succ 2 と Nat.succ 3 は同じコンストラクタから来ているという意味では同じ。そういう同一視をする関数 Nat → Bool を定義するにはパターンマッチをすれば良いが、では自動生成できるか？
コード例にするとこんな感じ:
```lean
inductive EnumLike : Type where
| foo (name : String)
…