search

pydatabangalore / talks

Talks at PyData Bangalore meetups
MIT License
36 stars 11 forks source link

Interpretability: Cracking open the Black Box #30

Closed manujosephv closed 4 years ago

manujosephv commented 4 years ago

Title

Interpretability: Cracking open the Black Box

Description

A review of Interpretable models, how to interpret them and common pitalls. And then move on to look at techniques to explain black box models and how to correctly interpret the results.

Duration

Audience

Talk is for intermediate as well as experienced ML practitioners. Prerequisites:

  1. Basic Knowldge of Machine Learning Algorithms
  2. Python (soft requirement as the examples are in Python, but the concepts are universal)

Outline

As the machine learning field matured and more and more models have started to be in use in the wild, the need for explainability has become more and more important. The talk would be structured as below

  • What is Interpretability?
  • Interpretable Models - How to interpret them?
  • Blackbox model explanation techniques
  • Mean Decrease in Impurity,
  • Permutation Importance
  • Partial Dependence Plots
  • LIME
  • Shapely Values and SHAP

Additional notes

https://www.linkedin.com/in/manujosephv/

TrigonaMinima commented 4 years ago

Hi @manujosephv thanks for the proposal.

NirantK commented 4 years ago

Brilliant topic set Manu! Sad that I'd have to miss this :(

On Tue, 7 Jan 2020, 12:54 Shivam Rana, notifications@github.com wrote:

Hi @manujosephv https://github.com/manujosephv thanks for the proposal.

  • Since we have a mixed crowd - beginners to advanced - it'd be good if you could add a little explanation to the models you're going to present.
  • Also it'd be good if you make it clear in the outline what kind of models (cv, nlp, traditional ml algos) you're going to target. If it's going to generic then i guess it's fine.
  • Would you be available for the talk on 18th Jan?

— You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub https://github.com/pydatabangalore/talks/issues/30?email_source=notifications&email_token=AAYZUPNVLVN6NEIP5VW5Q3DQ4QU4NA5CNFSM4KDSSM72YY3PNVWWK3TUL52HS4DFVREXG43VMVBW63LNMVXHJKTDN5WW2ZLOORPWSZGOEIH6YPQ#issuecomment-571468862, or unsubscribe https://github.com/notifications/unsubscribe-auth/AAYZUPP32XCXC5JFQ7WEYQ3Q4QU4NANCNFSM4KDSSM7Q .

manujosephv commented 4 years ago

The talk would be centered around Traditional ML Algos, although some of the techniques can be used for CV and NLP problems as well. Explanation Techniques - Mean Decrease in Impurity, Leave One Co-variate Out Importance, Permutation Importance, Partial Dependence Plots, ICE plots, Locally Interpretable Model Agnostic Explanations (LIME), Shapely Values

TrigonaMinima commented 4 years ago

Awesome! Your talk is final for the event. Putting you on the 3rd slot - 11:45 AM to 12:30 PM. See you there.

TrigonaMinima commented 4 years ago

Hey @manujosephv, can you please upload the ppt somewhere on cloud (google docs, google drive, slide share, github) and share the link here? Need to update the readme with the links to the slides.

manujosephv commented 4 years ago

@TrigonaMinima I've uploaded the ppt in this link https://drive.google.com/open?id=1GnpWyHXNNx-wkRgNFFvzz-iVE69ML-bV

@NirantK Even if you missed the session, if you can sit through a long three part blog, you can get the information here... https://deep-and-shallow.com/2019/11/13/interpretability-cracking-open-the-black-box-part-i/ https://deep-and-shallow.com/2019/11/16/interpretability-cracking-open-the-black-box-part-ii/ https://deep-and-shallow.com/2019/11/24/interpretability-cracking-open-the-black-box-part-iii/

TrigonaMinima commented 4 years ago

@manujosephv I dont think the ppt is publicly available. Please make it public.

manujosephv commented 4 years ago

Oops.. have made it public..

On Thu, Jan 23, 2020, 7:05 PM Shivam Rana notifications@github.com wrote:

@manujosephv https://github.com/manujosephv I dont think the ppt is publicly available. Please make it public.

— You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub https://github.com/pydatabangalore/talks/issues/30?email_source=notifications&email_token=ACQFRTIG6KOS3JK6F5M5C73Q7GMK5A5CNFSM4KDSSM72YY3PNVWWK3TUL52HS4DFVREXG43VMVBW63LNMVXHJKTDN5WW2ZLOORPWSZGOEJXMDFY#issuecomment-577683863, or unsubscribe https://github.com/notifications/unsubscribe-auth/ACQFRTMXFMAMA3FBQ25TEYDQ7GMK5ANCNFSM4KDSSM7Q .

TrigonaMinima commented 4 years ago

Yeah i can see it now. Thanks. See you in the future talks.