Skip to content

Instantly share code, notes, and snippets.

Embed
What would you like to do?
Etaler_API_example.ipynb
Sorry, something went wrong. Reload?
Sorry, we cannot display this file.
Sorry, this file is invalid so it cannot be displayed.
@alior101
Copy link

alior101 commented Aug 4, 2019

Awesome work Martin! Can't wait to start playing with it :)

@cyberslam
Copy link

cyberslam commented Aug 4, 2019

Hi Martin,
I had trouble signing in CERN to use ROOT, using
my Google account. Didn't try Colab since you said it won't
work with ROOT.

         I also try in my MATE terminal (Ubuntu 18.04) using 

your suggested command but got error:

sfl@T500:~$ root --notebook

Command 'root' not found, did you mean:

command 'toot' from snap toot (0.21.0)
command 'roo' from snap roo (2.0.3)
command 'rootv' from deb xawtv
command 'rott' from deb rott
command 'proot' from deb proot

See 'snap info ' for additional versions.

sfl@T500:~$

       Anyway, I read thru your "Etaler_API_example.ipynb" and indeed

it is impressive!!! You are a brilliant programmer and I can hardly
wait for the Python Wrapper for your c++ notebook. :))

      I noticed there were some discrepancies between Etaler's

graph outputs as compared with nupic. Etaler has too many
changes from nupic that it will be difficult for you to troubleshoot what
caused the changes. You should just overlay both Etaler's and
Nupic outputs together in one graph if you can: starting from the decoder
front end, and see how it affects the final outputs.

      Thanks Martin.

Samuel

@marty1885
Copy link
Author

marty1885 commented Aug 5, 2019

your suggested command but got error:

ROOT is not installed by default on Ubuntu based systems. Nor is it is the official repository. I installed it via Arch Linux's repo (I run Arch)

sudo pamcan -S root

You'll have to build ROOT yourself if you are on Ubuntu/Debian. @alior101, have you got a minute? I'm not sure how to build ROOT on Ubuntu/Debian.

@marty1885
Copy link
Author

marty1885 commented Aug 5, 2019

Etaler has too many changes from nupic that it will be difficult for you to troubleshoot what caused the changes.

There are some changes I know is causing the differences.

First, Etaler implements the original TM algorithm back from 2011 and updating to the new connect-to-all method. While NuPIC uses the updated TM algorithm from 2016 (but still staying on the old connect-once method). Secondly, I didn't set the hyper parameters in Etaler to what NuPIC has. Third, NuPIC uses anomaly likelihood while Etaler uses the raw anomaly score (as I haven't implement that yet)

@alior101
Copy link

alior101 commented Aug 5, 2019

@cyberslam
Copy link

cyberslam commented Aug 5, 2019

@marty1885
Copy link
Author

marty1885 commented Aug 5, 2019

@cyberslam You might want to check your email settings and edit your post. The format os way off and other data is leaking trough.

@cyberslam
Copy link

cyberslam commented Aug 5, 2019

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment