Did the Model
Understand
the Question?


TensorFlow & Deep Learning SG


Martin Andrews @ reddragon.ai

4 September 2018

About Me

  • Machine Intelligence / Startups / Finance
    • Moved from NYC to Singapore in Sep-2013
  • 2014 = 'fun' :
    • Machine Learning, Deep Learning, NLP
    • Robots, drones
  • Since 2015 = 'serious' :: NLP + deep learning
    • & GDE ML; TF&DL co-organiser
    • & Papers...
    • & Dev Course...

About Red Dragon AI

  • Google Partner : Deep Learning Consulting & Prototyping
  • SGInnovate/Govt : Education / Training
  • Products :
    • Conversational Computing
    • Natural Voice Generation - multiple languages
    • Knowledgebase interaction & reasoning

Outline

  • whoami = DONE
  • What is VQA ?
  • VQA models
  • "Did the Model Understand the Question?"
  • Demo of VQA ...
  • ... showing some limitations
  • Wrap-up

VQA

VQA Samples

VQA samples

VQA Questions

VQA questions

VQA Network

  • Example network described in original paper
  • ~57% accuracy
VQA network

VQA Modern Network

VQA Show Attend and Tell network

Paper Time

Paper Idea

  • VQA performance looks great!
    • But is the model doing what we think it is?
  • Test by changing the question :
    • Remove words
    • Add functionless phrases
  • Similiar to adversarial attacks

Surprises

  • If only question word allowed is "color" ...
    • model still more than 50% of original accuracy
  • If all questions are *empty* ...
    • model still more than 44% of original accuracy
  • More generally : Model doesn't care much about nouns

Distractors

  • These phrases can kill performance :
    • In not a lot of words, ...
    • In not many words, ...
    • What is the answer to : ...
  • But these phrases do no harm :
    • Answer this : ...
    • Answer this for me : ...
    • Tell me ...

Demo

  • Whole notebook should auto-gather code + data


https://github.com/
mdda/deep-learning-workshop/
notebooks/work-in-progress/
2018-08_DidTheModelUnderstandTheQuestion/
VQA_playground.ipynb



Load Directly into Colab

Wrap-up

  • Black-box models can do cool stuff
  • Good idea to test their limitations
  • Hope that Explainability theme was useful
GitHub - mdda

* Please add a star... *

Deep Learning
MeetUp Group

Deep Learning
Developer Course

  • JumpStart module is Module #1 of 5
  • Each 'module' will include :
    • Instruction
    • Individual Projects
    • 70%-100% funding via IMDA for SG/PR
  • 1-2 Nov : Advanced Computer Vision
  • 22-23 Nov : Advanced NLP
  • Location : SGInnovate

Deep Learning : Jump-Start Workshop

RedDragon AI
Intern Hunt

  • Opportunity to do Deep Learning all day
  • Work on something cutting-edge
  • Location : Singapore
  • Status : SG/PR FTW
  • Need to coordinate timing...

- QUESTIONS -


Martin.Andrews @
RedCatLabs.com

Martin.Andrews @
RedDragon.AI


My blog : http://blog.mdda.net/

GitHub : mdda