dice-embeddings

dice-embeddings copied to clipboard

We should integrate [RiemannianSGD](https://github.com/ibalazevic/multirelational-poincare/blob/master/rsgd.py). Update: Benchmark MurP performance with and without RiemannianSGD

Demirrr

Kolmogorov-Arnold Networks (KANs) over triple embeddings

3

Integrate Kolmogorov-Arnold Networks (KANs) model for forward_triples KAN(s,p,o) -> [0,1] forward triples KAN(s,p) -> [0,....,] KvsAll

Demirrr

Expanding to other datasets

1

@Demirrr Any interest in creating pretrained model checkpoints for other large-scale graphs? In particular, aside from YAGO/WN18RR/FB15K-237, there are several other graphs that are used for alignment or relation extraction...

akalino

Combining Neural Link Predictor, Numerical Literal Predictor under LitCQD,

2

Inputs 1. Train a neural link predictor (e.g. [KECI](https://link.springer.com/chapter/10.1007/978-3-031-43418-1_34)) 2. Numerical literal predictor (e.g. [LiteralE](https://arxiv.org/abs/1802.00934) or design a new one with us :) ) Combine (1) and (2) under [LitCQD](https://arxiv.org/abs/2304.14742)...

Demirrr

Logic Tensor Networks (LTN)

1

Integrate https://github.com/logictensornetworks/logictensornetworks LTN can be particularly usefull for multi-hop query answering

Demirrr

Demirrr

Using tqdm in the cpu train setup

7

If `--trainer torchCPUTrainer`, https://github.com/dice-group/dice-embeddings/blob/0849d2c46ea4bf06d07db6f3d0d2ad01fc230af5/dicee/scripts/run.py#L56 then TorchTrainer is initialized to train an embedding model. https://github.com/dice-group/dice-embeddings/blob/0849d2c46ea4bf06d07db6f3d0d2ad01fc230af5/dicee/trainer/torch_trainer.py#L9 Yet, TorchTrainer does not use a progress bar. We should integrate https://github.com/tqdm/tqdm into the TorchTrainer that...

Demirrr

Updating the README to show the test coverage and editing the documents to show the test coverge in detail as we have done for [ontolearn](https://ontolearn-docs-dice-group.netlify.app/usage/09_further_resources#code-coverage) would be very good. Could...

Demirrr

We may want to include https://github.com/dice-group/embeddings.cc diretctly so that by providing the path of a pretrained model. Embeddings can be diretly used

Demirrr

When training an embedding model on a KG, I am getting the following error stack: ```bash Reading with pandas.read_csv with sep ** s+ ** ... Traceback (most recent call last):...

sshivam95