addressalign-toparrow-leftarrow-rightbackbellblockcalendarcameraccwcheckchevron-downchevron-leftchevron-rightchevron-small-downchevron-small-leftchevron-small-rightchevron-small-upchevron-upcircle-with-checkcircle-with-crosscircle-with-pluscrossdots-three-verticaleditemptyheartexporteye-with-lineeyefacebookfolderfullheartglobegmailgooglegroupsimageimagesinstagramlinklocation-pinm-swarmSearchmailmessagesminusmoremuplabelShape 3 + Rectangle 1outlookpersonJoin Group on CardStartprice-ribbonShapeShapeShapeImported LayersImported LayersImported Layersshieldstartickettrashtriangle-downtriangle-uptwitteruserwarningyahoo

#dsGhent: Datatonic teaches Deep Learning 1/4

Update: Datatonic is bringing some extra people to help out in teaching; 100 people can come to the sessions!


Hi everyone,

we're glad to announce that we're providing Google's Deep Learning course on Udacity and taught by team Datatonic. Here are some details:

Session 1 , Thursday 1 December 2016:
- Technical introduction to TensorFlow
- Setting up your TensorFlow environment
- Udacity Session 1: From Machine Learning to Deep Learning + assignment

Session 2, Wednesday 7 December 2016:
- Udacity Session 2: Deep Neural Networks


Session 3, Thursday 15 December 2016: 
- Udacity Session 3: Convolutional Neural Networks


Session 4, Thursday 22 December 2016:
- Udacity Session 4: Deep Models for text and sequences


Here's some info on who'll be teaching these sessions:

Datatonic
Datatonic is a team of data science experts that help corporations unleash the power of data. They use Google Cloud Platform, data visualization technologies (like Tableau or Spotfire) and machine learning to build breakthrough solutions. Either as expert advisors, or built as a fully managed solution with support end-to-end (A3S).


Hope to see you there!

Join or login to comment.

  • Dimitri N.

    Thanks for the very interesting course!
    I have some troubles connecting to the Google VM again and so I can't open my Python Notebook. Is this something trivial and anybody that can help please?
    Thanks!

    3 days ago

    • Matthias F.

      Hi Dimitri, I assume you tried to connect to the notebook after restarting it?
      Two required steps:
      1) After restarting the machine, you should restart the notebook server on the machine:
      - SSH to the machine and paste/run "jupyter notebook --no-browser --ip=[masked] --port=80 & disown"
      2) The machine might have a new public IP address, so be sure you are using the new public IP address of the machine.

      1 · 2 days ago

  • Niek B.

    Thanks for organizing this course!

    I had some trouble understanding *why* all of these these magical equations work so well and found this article to give a very conceptual and visual explanation: https://cloud.google.com/blog/big-data/2016/07/understanding-neural-networks-with-tensorflow-playground

    2 · 5 days ago

    • Hendrik D.

      You can run it yourself and add features, such as 'angle' and 'radius' by running the github repo and editing 'INPUTS' in this file: https://github.com/ten...­ - should make the spiral one train much better

      1 · 4 days ago

  • Pieter B.

    Thanks @teamdatatonic for the session yesterday!

    For those of you who reached Problem 6, try out a Random Forest and compare its performance to Logistic Regression.

    Here is some quick and dirty code to help you out:
    import sklearn.ensemble
    n = 5000
    model = sklearn.ensemble.RandomForestClassifier(n_estimators=100)
    model.fit(train_dataset[:n].reshape((-1, 28*28)), train_labels[:n])
    print(model.score(valid_dataset.reshape((-1, 28*28)), valid_labels), model.score(test_dataset.reshape((-1, 28*28)), test_labels))

    As you will see Random Forests perform better on this task than Logistic Regression. Which is why I didn't cover it in the ML101++ session. But don't worry Deep Learning works even better...

    6 · 6 days ago

    • François D.

      Did you really get a significant speed improvement with this? I tried %time on the fit, both with and without the n_jobs argument and it didn't change much on my old PC (it does have 4 cores) ~ 5 sec in both cases :-) But maybe this is due to the fact that I'm using Docker Toolbox which doesn't know how many cores are available?

      5 days ago

    • Pieter B.

      If you set n_jobs=-1 you will use all your cores/hyperthreads in your machine and not just 4.

      But for some reason it stopped working for me since I updated my Anaconda so also only single core execution for me as well...

      5 days ago

  • Mathieu H.

    Where/How do we hand in our notebooks?

    6 days ago

    • Hendrik D.

      we'll talk about this next session :)

      2 · 5 days ago

  • Wietse H.

    This is free of charge?

    November 21

Our Sponsors

  • yes you can sponsor

    If you want to sponsor our activities please contact Philippe Van Impe

  • O'Reilly

    Special discounts when you use discount code UGBDSC

People in this
Meetup are also in:

Sign up

Meetup members, Log in

By clicking "Sign up" or "Sign up using Facebook", you confirm that you accept our Terms of Service & Privacy Policy