I'm trying to compute the embedding of a knowledge graph. My first approach involves using node2vec, but I have two issues: How to select the hyperparameters , like p , q , num_walks and dimensions . Currently, I'm training hundreds of node2vec models with different combinations of these parameters and then evaluating models based on training and testing embedding data, but I was wondering if there was a more efficient way to do that. This leads us to... How to run node2vec more efficiently for hyperparameter tuning (in case there is no better way than the one introduced in point 1). I've tried nodevector's version but it still needs to tune parameters.

How to tune node2vec hyperparameters

I'm trying to compute the embedding of a knowledge graph. My first approach involves using node2vec, but I have two issues:

How to select the hyperparameters, like p, q, num_walks and dimensions. Currently, I'm training hundreds of node2vec models with different combinations of these parameters and then evaluating models based on training and testing embedding data, but I was wondering if there was a more efficient way to do that. This leads us to...
How to run node2vec more efficiently for hyperparameter tuning (in case there is no better way than the one introduced in point 1). I've tried nodevector's version but it still needs to tune parameters.