Thank you for the package

[scienceML]

Would love to see your examples. and could not find py.test - was it part of the package?
your base module for windows does not have it

All the best!

[scienceML]

it seems like your package does not work even with your own example. it's pity

[dareneiri]

@scienceaiden. I have been able to use this package without any issue. I admit the documentation can be improved, but I've used publicly available survival datasets and get CI's greater than 0.6. I've commented on previous issues where I document how I've been able to use DeepSurv. Keep in mind that if you get deepsurv from PyPi (pip) then you'll be getting an older version than what's available from the master branch.

[scienceML]

@dareneiri i am happy for you.
I have been trying to download using pip or running directly from source code to no avails (did not see difference in code). and why do you think 0.6 is good result? theano proven to have much higher results.

[dareneiri]

@scienceaiden My point regarding getting a CI >0.6 was that DeepSurv works. For datasets I'm working with I usually get 0.8-0.9.

Can't say I understand why you're getting the same code from pip or from github. They're different for me. If I download from https://pypi.python.org/pypi/deepsurv/0.1.0 and check the files there, the source code is completely different from what's available on the master branch.

If you provide specific details about how you're trying to install and get DeepSurv to work, then I can try to help (I am not involved with DeepSurv directly, just very familiar with it).

[scienceML]

@dareneiri thank you for the help.

you are right - if i compare master ( in https://github.com/jaredleekatzman/DeepSurv) with deepsurv (in anaconda site-packages) folder, they are different.
what i meant is that i take all those class methods and definition and embed in the code and run, it gave me all sort of errors starting with different shape, etc. etc.

please guide me how can i install from the master properly.

[dareneiri]

If you haven't already, I recommend using a virtual environment to play with DeepSurv. This way you also start fresh and avoid any conflicts with your current python installation.

For example, with Anaconda, create a new virtual environment.

Then activate that virtual environment, and install Theano/Lasange using the latest bleeding edge versions:

pip install --upgrade https://github.com/Theano/Theano/archive/master.zip
pip install --upgrade https://github.com/Lasagne/Lasagne/archive/master.zip

Then install Deepsurv:

pip install git+https://github.com/jaredleekatzman/DeepSurv.git

Run python and make sure you can import theano/Deepsurv without any errors.
You can try using survival data from public sources, like the Haberman dataset. Note that since this is a NN, you'll need to normalize your data so that feature values are between 0-1.

Hope this brief writeup helps.

[scienceML]

@dareneiri thanks for the awesome writeup

i did like you said and when i import theano, i have this error message about lazylinker_ext.pyd Access is denied.

have you had experienced before?

[dareneiri]

Looks like it might be related to Windows updates?
Theano/Theano#2733

But I can't say for sure as I haven't had that issue; I installed theano on a linux system.

[scienceML]

@dareneiri thank you - you helped enough

btw: cannot get access to git+https://github.com/jaredleekatzman/DeepSurv.git now...

[dareneiri]

what is specific error?

[scienceML]

@dareneiri access denied

[dareneiri]

I don't have a problem running this on my system:
pip install git+https://github.com/jaredleekatzman/DeepSurv.git
This is likely coming from your own system and its own permission settings for your directory.

[scienceML]

@dareneiri ok, i will try to reboot router just in case as well.

thank you for all your help

[scienceML]

@dareneiri , after reinstalling everything from scratch, it works. thank you

[arturomoncadatorres]

As of today, I managed to install DeepSurv in a Windows 10 machine, 64 bit, running Python 3.7 and Anaconda using dareneiri instructions. However, I still needed to install h5py. For the sake of completion, this is what worked for me:

pip install --upgrade https://github.com/Theano/Theano/archive/master.zip
pip install --upgrade https://github.com/Lasagne/Lasagne/archive/master.zip
pip install git+https://github.com/jaredleekatzman/DeepSurv.git
pip install h5py