Estimate run time

[vforget]

Hi,

Is there a way to estimate run time of biglasso()? I have a data set of ~341k individuals vs 640k SNPs. This took a days to run, resulting in coefficients for 45 lambdas. I have another data set that was started around the same time that consists of 341k individuals and 332k SNPs. It is on lambda 65 and still running.

So:

1. Is there a way to determine how many lambdas biglasso() will test? Particularly if I already know at what lambda produces the min RMSE and other parameters I put into the training (dfmax etc)?
2. If possible, a message when biglasso() starts processing that reports how many lambda it will test would be a useful addition.

Thanks again for such a useful R package!

Vince

[YaohuiZeng]

@vforget Really sorry for so late response.

Regarding your questions:

1. by default, 100 lambdas were pre-computed internally. This is explained in the doc. On the other hand, you can specify your own sequence of lambda values via the parameter lambda, which also explained in the doc.

2. biglasso function has a verbose argument that does what you want.

Hope that helps. Let me know if there are other questions. Thanks!