Rgtsvm::svm running slower than e1071::svm

[NathanSkene]

Got Rgtsvm installed and tested it out on a multi class classification task. Training dataset has 12713 samples with 26 variables for each. Rgtsvm is running 1/3rd slower than e1071. Any guesses on what could be causing this? Happy to send over the dataset if that might help. Thanks!

> start.time <- Sys.time()
> model_svm1 <- Rgtsvm::svm(Celltype ~ . , trainingData)
> end.time <- Sys.time()
> time.taken <- end.time - start.time
> time.taken
Time difference of 1.031099 mins
>
> start.time <- Sys.time()
> model_svm2 <- e1071::svm(Celltype ~ . , trainingData)
> end.time <- Sys.time()
> time.taken <- end.time - start.time
> time.taken
Time difference of 43.36015 secs

[NathanSkene]

Link to the dataset is here: https://imperiallondon-my.sharepoint.com/:u:/g/personal/nskene_ic_ac_uk/EX-_0OGm2e1Luhn9E1-9Wm0BYlYP4DCwCSQNEAVTwLd33g?e=1fj3y9

Password is "trainingData"

[wzhy2000]

I got the same results using your data. I think your data is too small to be used to show GPU's performance. Would you like to try the big data, for example, 500K or 1M samples?