Issue about the results of function vis()
gaosh opened this issue · 7 comments
Firstly, thanks for the nice implementation.
I tried this algorithm on the cifar 10 dataset. The problem is that I got different results from the same input. Is there any method to deal with it, and make the results more stable?
In the first place, I'd strongly encourage you to use the version in branch reference (which hopefully will be merged into and become branch 0.1.6 tonight...) A lot of bugs have been fixed.
The new branch uses a difference random number generator where its possible to set the random seed properly, but I haven't written code to do that yet. I may try to get to it this week, although if you wanted to make a PR, what needs to be changed is pretty straightforward and it wouldn't take you very long.
I'm really eager to see what the results are of running this on cifar!
Thanks for the reply. There are the results comes from 256 dimension
features of cifar10 test set.
On Wed, Aug 3, 2016 at 12:41 AM, elbamos notifications@github.com wrote:
I'm really eager to see what the results are of running this on cifar!
—
You are receiving this because you authored the thread.
Reply to this email directly, view it on GitHub
#14 (comment),
or mute the thread
https://github.com/notifications/unsubscribe-auth/AQSH_tpDUyATEihG7GC-zWd5kS5W7op0ks5qcBwSgaJpZM4Ja-OK
.
@gaosh Branch 0.1.7, which I've just pushed, lets you set a random seed for reproducibility.
Note that setting the seed will impact performance, because it is also necessary to turn off multithreading or the results would be non-deterministic.
Ok, I will try this new version. Thanks a lot.
2016年8月7日星期日，elbamos notifications@github.com 写道：
@gaosh https://github.com/gaosh Branch 0.1.7, which I've just pushed,
lets you set a random seed for reproducibility.Note that setting the seed will impact performance, because it is also
necessary to turn off multithreading or the results would be
non-deterministic.—
You are receiving this because you were mentioned.
Reply to this email directly, view it on GitHub
#14 (comment),
or mute the thread
https://github.com/notifications/unsubscribe-auth/AQSH_kK68-mfk0N--123qPDq8xKPxCkeks5qdkTLgaJpZM4Ja-OK
.
Yes, I want to close the issue.
2016年8月18日星期四，elbamos notifications@github.com 写道：
@gaosh https://github.com/gaosh If 0.1.7 worked, do you want to close
this issue?—
You are receiving this because you were mentioned.
Reply to this email directly, view it on GitHub
#14 (comment),
or mute the thread
https://github.com/notifications/unsubscribe-auth/AQSH_uW9xFWpfyKw3oJ27uD8EA4XBvfUks5qg-PugaJpZM4Ja-OK
.