SlideShare a Scribd company logo
Interpretability beyond feature attribution:
Testing with Concept Activation Vectors
TCAV
Been Kim
Presenting work with a lot of awesome people inside and outside of Google:
Marten Wattenberg, Julius Adebayo, Justin Gilmer, Carrie Cai, James Wexler,
Fernanda Viegas, Ian Goodfellow, Mortiz Hardt, Rory Sayres
My goal
interpretability
!2
To use machine learning responsibly
we need to ensure that
1. our values are aligned
2. our knowledge is reflected
for everyone.
http://blogs.teradata.com/
Machine
Learning
Models
Human
My goal
interpretability
!3
To use machine learning responsibly
we need to ensure that
1. our values are aligned
2. our knowledge is reflected
for everyone.
http://blogs.teradata.com/
Machine
Learning
Models
Human
ingredients for interpretability methods.
!4
?
Some quality function
Model
Model
Class1
Class0
Data
Class1
Class0
Data
Class1
Class0
Human
What’s ML?
If I were you, I
would train a
neural network.
newbie
expert
Class1
Class0
Task
• Local vs. global
• Simple explanations vs.
more complex but more
accurate explanations
• Low or high stake domains
Interpretability methods
Post-training explanations
Building inherently interpretable models
Explaining data
Agenda
Post-training explanations
Agenda
1. Revisiting existing methods:
Saliency maps
2. Making explanations using the way humans think:
Testing with concept activation vectors (TCAV)
Post-training explanations
Agenda
Sanity Checks for Saliency Maps
Joint work with Adebayo, Gilmer, Goodfellow, Hardt, [NIPS 18]
1. Revisiting existing methods:
Saliency maps
2. Making explanations using the way humans think:
Testing with concept activation vectors (TCAV)
Post-training explanations
Problem:
Post-training explanation
!15
cash-machine-ness
A trained
machine learning model
(e.g., neural network)
Why was this a
cash machine?
One of the most popular interpretability methods for images:
Saliency maps
!16
SmoothGrad [Smilkov, Thorat, K., Viégas, Wattenberg ’17]
Integrated gradient [Sundararajan, Taly, Yan ’17]
picture credit: @sayres
Caaaaan do! We’ve got
saliency maps to measure
importance of each pixel!
a logit
pixel i,j
One of the most popular interpretability methods for images:
Saliency maps
!17
SmoothGrad [Smilkov, Thorat, K., Viégas, Wattenberg ’17]
Integrated gradient [Sundararajan, Taly, Yan ’17]
widely used
for images
local
undestandingNN
humans’
subjective
judgement
picture credit: @sayres
One of the most popular interpretability methods for images:
Saliency maps
!18
Sanity check:
If I change M a lot, will human
perceive that E has changed a lot?
SmoothGrad [Smilkov, Thorat, K., Viégas, Wattenberg ’17]
Integrated gradient [Sundararajan, Taly, Yan ’17]
widely used
for images
local
undestandingNN
humans’
subjective
judgement
picture credit: @sayres
Some confusing behaviors of saliency maps.
Saliency map
Sanity Checks for Saliency Maps
Joint work with Adebayo, Gilmer, Goodfellow, Hardt, [NIPS 18]
Some confusing behaviors of saliency maps.
Randomized weights!
Network now makes garbage prediction.
Saliency map
Sanity Checks for Saliency Maps
Joint work with Adebayo, Gilmer, Goodfellow, Hardt, [NIPS 18]
Some confusing behaviors of saliency maps.
Saliency map
Randomized weights!
Network now makes garbage prediction.
!!!!!???!?
Sanity Checks for Saliency Maps
Joint work with Adebayo, Gilmer, Goodfellow, Hardt, [NIPS 18]
Some saliency maps look similar
when we randomize the network
(= making the network completely useless)
Before After
Guided
Backprop
Integrated
Gradient
Sanity Checks for Saliency Maps
Joint work with Adebayo, Gilmer, Goodfellow, Hardt, [NIPS 18]
• Potential human confirmation bias: Just because it
“makes sense” to humans, doesn’t mean they reflect
evidence for the prediction.
• Our discovery is consistent with other findings.
[Nie, Zhang, Patel ’18] [Ulyanov, Vedaldi, Lempitsky ’18]
• Some of these methods have been shown to be useful for
humans. Why? More studies needed.
What can we learn from this?
Sanity Checks for Saliency Maps
Joint work with Adebayo, Gilmer, Goodfellow, Hardt, [NIPS 18]
!24
local
undestanding
lay
person?
human’s
subjective
judgement
What can we do better?
Creating a wishlist.
Using
input
features
as a language
!25
local
undestanding
lay
person?
human’s
subjective
judgement
global
!
quantitive
Using
input
features
as a language
Something more human-friendly?
What can we do better?
Creating a wishlist.
Agenda
TCAV [ICML’18]
Joint work with Wattenberg, Gilmer, Cai, Wexler, Viegas, Sayres
1. Revisiting existing methods:
Saliency maps
2. Making explanations using the way humans think:
Testing with concept activation vectors (TCAV)
Post-training explanations
Problem:
Post-training explanation
!27
cash-machine-ness
A trained
machine learning model
(e.g., neural network)
Why was this a
cash machine?
TCAV [ICML’18]
Joint work with Wattenberg, Gilmer, Cai, Wexler, Viegas, Sayres
prediction:
Cash machine
https://pair-code.github.io/saliency/
SmoothGrad [Smilkov, Thorat, K., Viégas, Wattenberg ’17]
Common solution: Saliency map
!28
Let’s use this to help us
think about what what we
really want to ask.
prediction:
Cash machine
https://pair-code.github.io/saliency/
SmoothGrad [Smilkov, Thorat, K., Viégas, Wattenberg ’17]
What we really want to ask…
!29
Were there more pixels on the cash
machine than on the person?
Did the ‘human’ concept matter?
Did the ‘wheels’ concept matter?
prediction:
Cash machine
https://pair-code.github.io/saliency/
SmoothGrad [Smilkov, Thorat, K., Viégas, Wattenberg ’17]
What we really want to ask…
!30
Were there more pixels on the cash
machine than on the person?
Which concept mattered more?
Is this true for all other cash
machine predictions?
Did the ‘human’ concept matter?
Did the ‘wheels’ concept matter?
prediction:
Cash machine
https://pair-code.github.io/saliency/
SmoothGrad [Smilkov, Thorat, K., Viégas, Wattenberg ’17]
What we really want to ask…
!31
Oh no! I can’t express these concepts
as pixels!!
They weren’t my input features either!
Were there more pixels on the cash
machine than on the person?
Which concept mattered more?
Is this true for all other cash
machine predictions?
Did the ‘human’ concept matter?
Did the ‘wheels’ concept matter?
prediction:
Cash machine
https://pair-code.github.io/saliency/
SmoothGrad [Smilkov, Thorat, K., Viégas, Wattenberg ’17]
What we really want to ask…
!32
Were there more pixels on the cash
machine than on the person?
Which concept mattered more?
Is this true for all other cash
machine predictions?
Wouldn’t it be great if we can
quantitatively measure how
important any of these
user-chosen concepts are?
Did the ‘human’ concept matter?
Did the ‘wheels’ concept matter?
Quantitative explanation: how much a concept (e.g., gender, race)
was important for a prediction in a trained model.
…even if the concept was not part of the training.
Goal of TCAV:
Testing with Concept Activation Vectors
!33 ICML 2018
Goal of TCAV:
Testing with Concept Activation Vectors
!34
Doctor-ness
A trained
machine learning model
(e.g., neural network)
vactruth.com healthcommunitiesproviderservices
Goal of TCAV:
Testing with Concept Activation Vectors
!35
Doctor-ness
A trained
machine learning model
(e.g., neural network)
Was gender concept important
to this doctor image classifier?
vactruth.com healthcommunitiesproviderservices
Goal of TCAV:
Testing with Concept Activation Vectors
!36
Doctor-ness
TCAV score for
womennot women
Doctor
A trained
machine learning model
(e.g., neural network)
vactruth.com healthcommunitiesproviderservices
Was gender concept important
to this doctor image classifier?
Goal of TCAV:
Testing with Concept Activation Vectors
!37
Doctor-ness
TCAV score for
womennot women
Doctor
A trained
machine learning model
(e.g., neural network)
vactruth.com healthcommunitiesproviderservices
Was gender concept important
to this doctor image classifier?
TCAV provides
quantitative importance of
a concept if and only if your
network learned about it.
Goal of TCAV:
Testing with Concept Activation Vectors
!38
zebra-ness
A trained
machine learning model
(e.g., neural network)
Was striped concept important
to this zebra image classifier?
TCAV score for
not stripedstriped
Zebra
TCAV provides
quantitative importance of
a concept if and only if your
network learned about it.
TCAV
TCAV:
Testing with Concept Activation Vectors
!39
zebra-ness
A trained
machine learning model
(e.g., neural network)
Was striped concept important
to this zebra image classifier?
1. Learning CAVs
1. How to define
concepts?
Defining concept activation vector (CAV)
Inputs:
!40
Random
images
Examples of
concepts
A trained network under investigation
and
Internal tensors
!41
Inputs:
Train a linear classifier to
separate activations.
CAV ( ) is the vector
orthogonal to the decision
boundary.
[Smilkov ’17, Bolukbasi ’16 , Schmidt ’15]
Defining concept activation vector (CAV)
TCAV
TCAV:
Testing with Concept Activation Vectors
!42
zebra-ness
A trained
machine learning model
(e.g., neural network)
Was striped concept important
to this zebra image classifier?
1. Learning CAVs 2. Getting TCAV score
2. How are the CAVs
useful to get
explanations?
TCAV core idea:
Derivative with CAV to get prediction sensitivity
!43
TCAV score
Directional derivative with CAV
striped CAV
TCAV score
Directional derivative with CAV
TCAV core idea:
Derivative with CAV to get prediction sensitivity
!44
TCAV
TCAV:
Testing with Concept Activation Vectors
!45
zebra-ness
A trained
machine learning model
(e.g., neural network)
Was striped concept important
to this zebra image classifier?
1. Learning CAVs 2. Getting TCAV score
TCAV
TCAV:
Testing with Concept Activation Vectors
!46
zebra-ness
A trained
machine learning model
(e.g., neural network)
Was striped concept important
to this zebra image classifier?
1. Learning CAVs 2. Getting TCAV score 3. CAV validation
Qualitative
Quantitative
Quantitative validation:
Guarding against spurious CAV
Did my CAVs returned high sensitivity by chance?
!47
Learn many stripes CAVs
using different sets of
random images
Quantitative validation:
Guarding against spurious CAV
!48
……
Zebra
Quantitative validation:
Guarding against spurious CAV
!49
……
Zebra
Quantitative validation:
Guarding against spurious CAV
!50
Check the distribution of
is statistically
different from random
using t-test
TCAV score
random
……
Zebra
Quantitative validation:
Guarding against spurious CAV
!51
*
Recap TCAV:
Testing with Concept Activation Vectors
!52
1. Learning CAVs 2. Getting TCAV score 3. CAV validation
Qualitative
Quantitative
TCAV provides
quantitative importance of
a concept if and only if your
network learned about it.
Even if your training data wasn’t
tagged with the concept
Even if your input feature did
not include the concept
Results
1. Sanity check experiment
2. Biases in Inception V3 and GoogleNet
3. Domain expert confirmation from Diabetic Retinopathy
!53
Results
1. Sanity check experiment
2. Biases from Inception V3 and GoogleNet
3. Domain expert confirmation from Diabetic Retinopathy
!54
Sanity check experiment
!55
If we know the ground truth
(important concepts),
will TCAV match?
Sanity check experiment setup
!56
An image
+
Potentially noisy Caption
Sanity check experiment setup
!57
An image
+
Potentially noisy Caption
image
concept
models can use either
image or caption
concept for
classification.
caption
concept
Sanity check experiment setup
!58 Caption noise level in training set
An image
+
Potentially noisy Caption
image
concept
models can use either
image or caption
concept for
classification.
caption
concept
Sanity check experiment setup
!59
Test accuracy
with
no caption image
=
Importance of
image concept
Caption noise level in training set
image
concept
caption
concept
models can use either
image or caption
concept for
classification.
Sanity check experiment
!60
Caption noise level in training set Caption noise level in training set
Test accuracy
with
no caption image
!61
Cool, cool.
Can saliency maps do this too?
Can saliency maps communicate
the same information?
!62
Ground truth
Image
concept
Image
concept
Image
concept
Image
concept
Image
with caption
Human subject experiment:
Can saliency maps communicate the same
information?
• 50 turkers are
• asked to judge importance of
image vs. c. ept given saliency
maps.
• asked to indicate their confidence
• shown 3 classes (cab, zebra,
cucumber) x 2 saliency maps for
one model
!63
image caption
!64
• Random chance: 50%
• Human performance with
saliency map: 52%
• Humans can’t agree: more
than 50% no significant
consensus
• Humans are very confident
even when they are wrong.
Human subject experiment:
Can saliency maps communicate the same
information?
Human subject experiment:
Can saliency maps communicate the same
information?
• Random chance: 50%
• Human performance with
saliency map: 52%
• Humans can’t agree: more
than 50% no significant
consensus
• Humans are very confident
even when they are wrong.
!65
Results
1. Sanity check experiment
2. Biases from Inception V3 and GoogleNet
3. Domain expert confirmation from Diabetic Retinopathy
!66
TCAV in
Two widely used image prediction models
!67
TCAV in
Two widely used image prediction models
!68
Geographical
bias!
http://www.abc.net.au
TCAV in
Two widely used image prediction models
!69
Quantitative
confirmation to
previously
qualitative
findings
[Stock & Cisse,
2017]
Geographical
bias?
TCAV in
Two widely used image prediction models
!70
Quantitative
confirmation to
previously
qualitative
findings
[Stock & Cisse,
2017]
Geographical
bias?
Goal of interpretability:
To use machine learning responsibly
we need to ensure that
1. our values are aligned
2. our knowledge is reflected
Results
1. Sanity check experiment
2. Biases Inception V3 and GoogleNet
3. Domain expert confirmation from Diabetic Retinopathy
!71
Diabetic Retinopathy
• Treatable but sight-threatening conditions
• Have model to with accurate prediction of DR (85%)
[Krause et al., 2017]
!72
Concepts the ML model uses
Vs
Diagnostic Concepts human doctors use
Collect human doctor’s knowledge
!73
PRP
PRH/VH
NV/FP
VB
MA HMA
DR level 4
DR level 1
Concepts

belong to 

this level
Concepts do not 

belong to 

this level
TCAV for Diabetic Retinopathy
!74
PRP PRH/VH NV/FP VB
Green: domain expert’s label on concepts belong to the level
Red: domain expert’s label on concepts does not belong to the level
Prediction
class
DR level 4
Prediction
accuracy
High
Example
TCAV scores TCAV shows the
model is consistent
with doctor’s
knowledge when
model is accurate
PRP PRH/VH NV/FP VB
Green: domain expert’s label on concepts belong to the level
Red: domain expert’s label on concepts does not belong to the level
Prediction
class
DR level 4
Prediction
accuracy
High
Example
TCAV scores TCAV shows the
model is consistent
with doctor’s
knowledge when
model is accurate
TCAV shows the
model is inconsistent
with doctor’s
knowledge for classes
when model is less
accurate
DR level 1 Med
TCAV for Diabetic Retinopathy
!75
MA HMA
PRP PRH/VH NV/FP VB
Green: domain expert’s label on concepts belong to the level
Red: domain expert’s label on concepts does not belong to the level
Prediction
class
DR level 4
Prediction
accuracy
High
Example
TCAV scores TCAV shows the
model is consistent
with doctor’s
knowledge when
model is accurate
Level 1 was often confused to level 2.
DR level 1 Low
TCAV shows the
model is inconsistent
with doctor’s
knowledge for classes
when model is less
accurate
TCAV for Diabetic Retinopathy
!76
MA HMA
Goal of interpretability:
To use machine learning responsibly
we need to ensure that
1. our values are aligned
2. our knowledge is reflected
Summary:
Testing with Concept Activation Vectors
!77
stripes concept (score: 0.9)
was important to zebra class
for this trained network.
PRP PRH/VH NV/FP VB
Our values Our knowledge
TCAV provides
quantitative importance of
a concept if and only if your
network learned about it.
Joint work with Wattenberg, Gilmer, Cai, Wexler, Viegas, Sayres
ICML 2018

More Related Content

What's hot

論文紹介:DINO: DETR with Improved DeNoising Anchor Boxes for End-to-End Object D...
 論文紹介:DINO: DETR with Improved DeNoising Anchor Boxes for End-to-End Object D... 論文紹介:DINO: DETR with Improved DeNoising Anchor Boxes for End-to-End Object D...
論文紹介:DINO: DETR with Improved DeNoising Anchor Boxes for End-to-End Object D...
Toru Tamaki
 
【DL輪読会】Toward Fast and Stabilized GAN Training for Highfidelity Few-shot Imag...
【DL輪読会】Toward Fast and Stabilized GAN Training for Highfidelity Few-shot Imag...【DL輪読会】Toward Fast and Stabilized GAN Training for Highfidelity Few-shot Imag...
【DL輪読会】Toward Fast and Stabilized GAN Training for Highfidelity Few-shot Imag...
Deep Learning JP
 
[DL輪読会]ICLR2020の分布外検知速報
[DL輪読会]ICLR2020の分布外検知速報[DL輪読会]ICLR2020の分布外検知速報
[DL輪読会]ICLR2020の分布外検知速報
Deep Learning JP
 
Spatial Temporal Graph Convolutional Networks for Skeleton-Based Action Recog...
Spatial Temporal Graph Convolutional Networks for Skeleton-Based Action Recog...Spatial Temporal Graph Convolutional Networks for Skeleton-Based Action Recog...
Spatial Temporal Graph Convolutional Networks for Skeleton-Based Action Recog...
yukihiro domae
 
最新の異常検知手法(NIPS 2018)
最新の異常検知手法(NIPS 2018)最新の異常検知手法(NIPS 2018)
最新の異常検知手法(NIPS 2018)
ぱんいち すみもと
 
[Dl輪読会]dl hacks輪読
[Dl輪読会]dl hacks輪読[Dl輪読会]dl hacks輪読
[Dl輪読会]dl hacks輪読
Deep Learning JP
 
【論文読み会】Alias-Free Generative Adversarial Networks(StyleGAN3)
【論文読み会】Alias-Free Generative Adversarial Networks(StyleGAN3)【論文読み会】Alias-Free Generative Adversarial Networks(StyleGAN3)
【論文読み会】Alias-Free Generative Adversarial Networks(StyleGAN3)
ARISE analytics
 
【DL輪読会】Contrastive Learning as Goal-Conditioned Reinforcement Learning
【DL輪読会】Contrastive Learning as Goal-Conditioned Reinforcement Learning【DL輪読会】Contrastive Learning as Goal-Conditioned Reinforcement Learning
【DL輪読会】Contrastive Learning as Goal-Conditioned Reinforcement Learning
Deep Learning JP
 
[DL輪読会]Wasserstein GAN/Towards Principled Methods for Training Generative Adv...
[DL輪読会]Wasserstein GAN/Towards Principled Methods for Training Generative Adv...[DL輪読会]Wasserstein GAN/Towards Principled Methods for Training Generative Adv...
[DL輪読会]Wasserstein GAN/Towards Principled Methods for Training Generative Adv...
Deep Learning JP
 
SSII2021 [OS2-03] 自己教師あり学習における対照学習の基礎と応用
SSII2021 [OS2-03] 自己教師あり学習における対照学習の基礎と応用SSII2021 [OS2-03] 自己教師あり学習における対照学習の基礎と応用
SSII2021 [OS2-03] 自己教師あり学習における対照学習の基礎と応用
SSII
 
[DL輪読会]StarGAN: Unified Generative Adversarial Networks for Multi-Domain Ima...
 [DL輪読会]StarGAN: Unified Generative Adversarial Networks for Multi-Domain Ima... [DL輪読会]StarGAN: Unified Generative Adversarial Networks for Multi-Domain Ima...
[DL輪読会]StarGAN: Unified Generative Adversarial Networks for Multi-Domain Ima...
Deep Learning JP
 
StyleGAN解説 CVPR2019読み会@DeNA
StyleGAN解説 CVPR2019読み会@DeNAStyleGAN解説 CVPR2019読み会@DeNA
StyleGAN解説 CVPR2019読み会@DeNA
Kento Doi
 
【DL輪読会】Semi-Parametric Neural Image Synthesis
【DL輪読会】Semi-Parametric Neural Image Synthesis【DL輪読会】Semi-Parametric Neural Image Synthesis
【DL輪読会】Semi-Parametric Neural Image Synthesis
Deep Learning JP
 
Graph Attention Network
Graph Attention NetworkGraph Attention Network
Graph Attention Network
Takahiro Kubo
 
SSII2019TS: Shall We GANs?​ ~GANの基礎から最近の研究まで~
SSII2019TS: Shall We GANs?​ ~GANの基礎から最近の研究まで~SSII2019TS: Shall We GANs?​ ~GANの基礎から最近の研究まで~
SSII2019TS: Shall We GANs?​ ~GANの基礎から最近の研究まで~
SSII
 
深層学習による非滑らかな関数の推定
深層学習による非滑らかな関数の推定深層学習による非滑らかな関数の推定
深層学習による非滑らかな関数の推定
Masaaki Imaizumi
 
【DL輪読会】Efficiently Modeling Long Sequences with Structured State Spaces
【DL輪読会】Efficiently Modeling Long Sequences with Structured State Spaces【DL輪読会】Efficiently Modeling Long Sequences with Structured State Spaces
【DL輪読会】Efficiently Modeling Long Sequences with Structured State Spaces
Deep Learning JP
 
【論文読み会】Deep Clustering for Unsupervised Learning of Visual Features
【論文読み会】Deep Clustering for Unsupervised Learning of Visual Features【論文読み会】Deep Clustering for Unsupervised Learning of Visual Features
【論文読み会】Deep Clustering for Unsupervised Learning of Visual Features
ARISE analytics
 
[DL輪読会]ドメイン転移と不変表現に関するサーベイ
[DL輪読会]ドメイン転移と不変表現に関するサーベイ[DL輪読会]ドメイン転移と不変表現に関するサーベイ
[DL輪読会]ドメイン転移と不変表現に関するサーベイ
Deep Learning JP
 
BlackBox モデルの説明性・解釈性技術の実装
BlackBox モデルの説明性・解釈性技術の実装BlackBox モデルの説明性・解釈性技術の実装
BlackBox モデルの説明性・解釈性技術の実装
Deep Learning Lab(ディープラーニング・ラボ)
 

What's hot (20)

論文紹介:DINO: DETR with Improved DeNoising Anchor Boxes for End-to-End Object D...
 論文紹介:DINO: DETR with Improved DeNoising Anchor Boxes for End-to-End Object D... 論文紹介:DINO: DETR with Improved DeNoising Anchor Boxes for End-to-End Object D...
論文紹介:DINO: DETR with Improved DeNoising Anchor Boxes for End-to-End Object D...
 
【DL輪読会】Toward Fast and Stabilized GAN Training for Highfidelity Few-shot Imag...
【DL輪読会】Toward Fast and Stabilized GAN Training for Highfidelity Few-shot Imag...【DL輪読会】Toward Fast and Stabilized GAN Training for Highfidelity Few-shot Imag...
【DL輪読会】Toward Fast and Stabilized GAN Training for Highfidelity Few-shot Imag...
 
[DL輪読会]ICLR2020の分布外検知速報
[DL輪読会]ICLR2020の分布外検知速報[DL輪読会]ICLR2020の分布外検知速報
[DL輪読会]ICLR2020の分布外検知速報
 
Spatial Temporal Graph Convolutional Networks for Skeleton-Based Action Recog...
Spatial Temporal Graph Convolutional Networks for Skeleton-Based Action Recog...Spatial Temporal Graph Convolutional Networks for Skeleton-Based Action Recog...
Spatial Temporal Graph Convolutional Networks for Skeleton-Based Action Recog...
 
最新の異常検知手法(NIPS 2018)
最新の異常検知手法(NIPS 2018)最新の異常検知手法(NIPS 2018)
最新の異常検知手法(NIPS 2018)
 
[Dl輪読会]dl hacks輪読
[Dl輪読会]dl hacks輪読[Dl輪読会]dl hacks輪読
[Dl輪読会]dl hacks輪読
 
【論文読み会】Alias-Free Generative Adversarial Networks(StyleGAN3)
【論文読み会】Alias-Free Generative Adversarial Networks(StyleGAN3)【論文読み会】Alias-Free Generative Adversarial Networks(StyleGAN3)
【論文読み会】Alias-Free Generative Adversarial Networks(StyleGAN3)
 
【DL輪読会】Contrastive Learning as Goal-Conditioned Reinforcement Learning
【DL輪読会】Contrastive Learning as Goal-Conditioned Reinforcement Learning【DL輪読会】Contrastive Learning as Goal-Conditioned Reinforcement Learning
【DL輪読会】Contrastive Learning as Goal-Conditioned Reinforcement Learning
 
[DL輪読会]Wasserstein GAN/Towards Principled Methods for Training Generative Adv...
[DL輪読会]Wasserstein GAN/Towards Principled Methods for Training Generative Adv...[DL輪読会]Wasserstein GAN/Towards Principled Methods for Training Generative Adv...
[DL輪読会]Wasserstein GAN/Towards Principled Methods for Training Generative Adv...
 
SSII2021 [OS2-03] 自己教師あり学習における対照学習の基礎と応用
SSII2021 [OS2-03] 自己教師あり学習における対照学習の基礎と応用SSII2021 [OS2-03] 自己教師あり学習における対照学習の基礎と応用
SSII2021 [OS2-03] 自己教師あり学習における対照学習の基礎と応用
 
[DL輪読会]StarGAN: Unified Generative Adversarial Networks for Multi-Domain Ima...
 [DL輪読会]StarGAN: Unified Generative Adversarial Networks for Multi-Domain Ima... [DL輪読会]StarGAN: Unified Generative Adversarial Networks for Multi-Domain Ima...
[DL輪読会]StarGAN: Unified Generative Adversarial Networks for Multi-Domain Ima...
 
StyleGAN解説 CVPR2019読み会@DeNA
StyleGAN解説 CVPR2019読み会@DeNAStyleGAN解説 CVPR2019読み会@DeNA
StyleGAN解説 CVPR2019読み会@DeNA
 
【DL輪読会】Semi-Parametric Neural Image Synthesis
【DL輪読会】Semi-Parametric Neural Image Synthesis【DL輪読会】Semi-Parametric Neural Image Synthesis
【DL輪読会】Semi-Parametric Neural Image Synthesis
 
Graph Attention Network
Graph Attention NetworkGraph Attention Network
Graph Attention Network
 
SSII2019TS: Shall We GANs?​ ~GANの基礎から最近の研究まで~
SSII2019TS: Shall We GANs?​ ~GANの基礎から最近の研究まで~SSII2019TS: Shall We GANs?​ ~GANの基礎から最近の研究まで~
SSII2019TS: Shall We GANs?​ ~GANの基礎から最近の研究まで~
 
深層学習による非滑らかな関数の推定
深層学習による非滑らかな関数の推定深層学習による非滑らかな関数の推定
深層学習による非滑らかな関数の推定
 
【DL輪読会】Efficiently Modeling Long Sequences with Structured State Spaces
【DL輪読会】Efficiently Modeling Long Sequences with Structured State Spaces【DL輪読会】Efficiently Modeling Long Sequences with Structured State Spaces
【DL輪読会】Efficiently Modeling Long Sequences with Structured State Spaces
 
【論文読み会】Deep Clustering for Unsupervised Learning of Visual Features
【論文読み会】Deep Clustering for Unsupervised Learning of Visual Features【論文読み会】Deep Clustering for Unsupervised Learning of Visual Features
【論文読み会】Deep Clustering for Unsupervised Learning of Visual Features
 
[DL輪読会]ドメイン転移と不変表現に関するサーベイ
[DL輪読会]ドメイン転移と不変表現に関するサーベイ[DL輪読会]ドメイン転移と不変表現に関するサーベイ
[DL輪読会]ドメイン転移と不変表現に関するサーベイ
 
BlackBox モデルの説明性・解釈性技術の実装
BlackBox モデルの説明性・解釈性技術の実装BlackBox モデルの説明性・解釈性技術の実装
BlackBox モデルの説明性・解釈性技術の実装
 

Similar to Interpretability beyond feature attribution quantitative testing with concept activation vectors (tcav)

Machine Learning ICS 273A
Machine Learning ICS 273AMachine Learning ICS 273A
Machine Learning ICS 273Abutest
 
Machine Learning ICS 273A
Machine Learning ICS 273AMachine Learning ICS 273A
Machine Learning ICS 273Abutest
 
What are machines learning? How might that impact design?
What are machines learning? How might that impact design?What are machines learning? How might that impact design?
What are machines learning? How might that impact design?
Andreas Wolters
 
Knowledge graphs, meet Deep Learning
Knowledge graphs, meet Deep LearningKnowledge graphs, meet Deep Learning
Knowledge graphs, meet Deep Learning
Connected Data World
 
Data Mining the City 2019 - Week 1
Data Mining the City 2019 - Week 1Data Mining the City 2019 - Week 1
Data Mining the City 2019 - Week 1
Violet Whitney, LEED AP
 
The Mobile Frontier
The Mobile FrontierThe Mobile Frontier
The Mobile Frontier
Rachel Hinman
 
Introduction to the Artificial Intelligence and Computer Vision revolution
Introduction to the Artificial Intelligence and Computer Vision revolutionIntroduction to the Artificial Intelligence and Computer Vision revolution
Introduction to the Artificial Intelligence and Computer Vision revolution
Darian Frajberg
 
Hacking Predictive Modeling - RoadSec 2018
Hacking Predictive Modeling - RoadSec 2018Hacking Predictive Modeling - RoadSec 2018
Hacking Predictive Modeling - RoadSec 2018
HJ van Veen
 
Crowdsourcing Wisdom
Crowdsourcing WisdomCrowdsourcing Wisdom
Crowdsourcing Wisdom
Vantte
 
Voxxed Athens 2018 - UX design and back-ends: When the back-end meets the user
Voxxed Athens 2018 - UX design and back-ends: When the back-end meets the userVoxxed Athens 2018 - UX design and back-ends: When the back-end meets the user
Voxxed Athens 2018 - UX design and back-ends: When the back-end meets the user
Voxxed Athens
 
NeurIPS2023 Keynote: The Many Faces of Responsible AI.pdf
NeurIPS2023 Keynote: The Many Faces of Responsible AI.pdfNeurIPS2023 Keynote: The Many Faces of Responsible AI.pdf
NeurIPS2023 Keynote: The Many Faces of Responsible AI.pdf
Lora Aroyo
 
Ai lecture1 final
Ai lecture1 finalAi lecture1 final
Ai lecture1 final
Shivam Agrawal
 
Machine Learning - Challenges, Learnings & Opportunities
Machine Learning - Challenges, Learnings & OpportunitiesMachine Learning - Challenges, Learnings & Opportunities
Machine Learning - Challenges, Learnings & Opportunities
CodePolitan
 
Human-Centered AI: Scalable, Interactive Tools for Interpretation and Attribu...
Human-Centered AI: Scalable, Interactive Tools for Interpretation and Attribu...Human-Centered AI: Scalable, Interactive Tools for Interpretation and Attribu...
Human-Centered AI: Scalable, Interactive Tools for Interpretation and Attribu...
polochau
 
Design, AI, and "-isms"
Design, AI, and "-isms"Design, AI, and "-isms"
Design, AI, and "-isms"
Michael Zarro, Ph.D.
 
Safety Bot Guaranteed -- Shmoocon 2017
Safety Bot Guaranteed -- Shmoocon 2017Safety Bot Guaranteed -- Shmoocon 2017
Safety Bot Guaranteed -- Shmoocon 2017
Richard Seymour
 
Continuous Automated Testing - Cast conference workshop august 2014
Continuous Automated Testing - Cast conference workshop august 2014Continuous Automated Testing - Cast conference workshop august 2014
Continuous Automated Testing - Cast conference workshop august 2014
Noah Sussman
 
On Inherent Complexity of Computation, by Attila Szegedi
On Inherent Complexity of Computation, by Attila SzegediOn Inherent Complexity of Computation, by Attila Szegedi
On Inherent Complexity of Computation, by Attila Szegedi
ZeroTurnaround
 
A Friendly Introduction to Machine Learning
A Friendly Introduction to Machine LearningA Friendly Introduction to Machine Learning
A Friendly Introduction to Machine Learning
Haptik
 
Web science AI and IA
Web science AI and IAWeb science AI and IA
Web science AI and IA
Fabien Gandon
 

Similar to Interpretability beyond feature attribution quantitative testing with concept activation vectors (tcav) (20)

Machine Learning ICS 273A
Machine Learning ICS 273AMachine Learning ICS 273A
Machine Learning ICS 273A
 
Machine Learning ICS 273A
Machine Learning ICS 273AMachine Learning ICS 273A
Machine Learning ICS 273A
 
What are machines learning? How might that impact design?
What are machines learning? How might that impact design?What are machines learning? How might that impact design?
What are machines learning? How might that impact design?
 
Knowledge graphs, meet Deep Learning
Knowledge graphs, meet Deep LearningKnowledge graphs, meet Deep Learning
Knowledge graphs, meet Deep Learning
 
Data Mining the City 2019 - Week 1
Data Mining the City 2019 - Week 1Data Mining the City 2019 - Week 1
Data Mining the City 2019 - Week 1
 
The Mobile Frontier
The Mobile FrontierThe Mobile Frontier
The Mobile Frontier
 
Introduction to the Artificial Intelligence and Computer Vision revolution
Introduction to the Artificial Intelligence and Computer Vision revolutionIntroduction to the Artificial Intelligence and Computer Vision revolution
Introduction to the Artificial Intelligence and Computer Vision revolution
 
Hacking Predictive Modeling - RoadSec 2018
Hacking Predictive Modeling - RoadSec 2018Hacking Predictive Modeling - RoadSec 2018
Hacking Predictive Modeling - RoadSec 2018
 
Crowdsourcing Wisdom
Crowdsourcing WisdomCrowdsourcing Wisdom
Crowdsourcing Wisdom
 
Voxxed Athens 2018 - UX design and back-ends: When the back-end meets the user
Voxxed Athens 2018 - UX design and back-ends: When the back-end meets the userVoxxed Athens 2018 - UX design and back-ends: When the back-end meets the user
Voxxed Athens 2018 - UX design and back-ends: When the back-end meets the user
 
NeurIPS2023 Keynote: The Many Faces of Responsible AI.pdf
NeurIPS2023 Keynote: The Many Faces of Responsible AI.pdfNeurIPS2023 Keynote: The Many Faces of Responsible AI.pdf
NeurIPS2023 Keynote: The Many Faces of Responsible AI.pdf
 
Ai lecture1 final
Ai lecture1 finalAi lecture1 final
Ai lecture1 final
 
Machine Learning - Challenges, Learnings & Opportunities
Machine Learning - Challenges, Learnings & OpportunitiesMachine Learning - Challenges, Learnings & Opportunities
Machine Learning - Challenges, Learnings & Opportunities
 
Human-Centered AI: Scalable, Interactive Tools for Interpretation and Attribu...
Human-Centered AI: Scalable, Interactive Tools for Interpretation and Attribu...Human-Centered AI: Scalable, Interactive Tools for Interpretation and Attribu...
Human-Centered AI: Scalable, Interactive Tools for Interpretation and Attribu...
 
Design, AI, and "-isms"
Design, AI, and "-isms"Design, AI, and "-isms"
Design, AI, and "-isms"
 
Safety Bot Guaranteed -- Shmoocon 2017
Safety Bot Guaranteed -- Shmoocon 2017Safety Bot Guaranteed -- Shmoocon 2017
Safety Bot Guaranteed -- Shmoocon 2017
 
Continuous Automated Testing - Cast conference workshop august 2014
Continuous Automated Testing - Cast conference workshop august 2014Continuous Automated Testing - Cast conference workshop august 2014
Continuous Automated Testing - Cast conference workshop august 2014
 
On Inherent Complexity of Computation, by Attila Szegedi
On Inherent Complexity of Computation, by Attila SzegediOn Inherent Complexity of Computation, by Attila Szegedi
On Inherent Complexity of Computation, by Attila Szegedi
 
A Friendly Introduction to Machine Learning
A Friendly Introduction to Machine LearningA Friendly Introduction to Machine Learning
A Friendly Introduction to Machine Learning
 
Web science AI and IA
Web science AI and IAWeb science AI and IA
Web science AI and IA
 

More from MLconf

Jamila Smith-Loud - Understanding Human Impact: Social and Equity Assessments...
Jamila Smith-Loud - Understanding Human Impact: Social and Equity Assessments...Jamila Smith-Loud - Understanding Human Impact: Social and Equity Assessments...
Jamila Smith-Loud - Understanding Human Impact: Social and Equity Assessments...
MLconf
 
Ted Willke - The Brain’s Guide to Dealing with Context in Language Understanding
Ted Willke - The Brain’s Guide to Dealing with Context in Language UnderstandingTed Willke - The Brain’s Guide to Dealing with Context in Language Understanding
Ted Willke - The Brain’s Guide to Dealing with Context in Language Understanding
MLconf
 
Justin Armstrong - Applying Computer Vision to Reduce Contamination in the Re...
Justin Armstrong - Applying Computer Vision to Reduce Contamination in the Re...Justin Armstrong - Applying Computer Vision to Reduce Contamination in the Re...
Justin Armstrong - Applying Computer Vision to Reduce Contamination in the Re...
MLconf
 
Igor Markov - Quantum Computing: a Treasure Hunt, not a Gold Rush
Igor Markov - Quantum Computing: a Treasure Hunt, not a Gold RushIgor Markov - Quantum Computing: a Treasure Hunt, not a Gold Rush
Igor Markov - Quantum Computing: a Treasure Hunt, not a Gold Rush
MLconf
 
Josh Wills - Data Labeling as Religious Experience
Josh Wills - Data Labeling as Religious ExperienceJosh Wills - Data Labeling as Religious Experience
Josh Wills - Data Labeling as Religious Experience
MLconf
 
Vinay Prabhu - Project GaitNet: Ushering in the ImageNet moment for human Gai...
Vinay Prabhu - Project GaitNet: Ushering in the ImageNet moment for human Gai...Vinay Prabhu - Project GaitNet: Ushering in the ImageNet moment for human Gai...
Vinay Prabhu - Project GaitNet: Ushering in the ImageNet moment for human Gai...
MLconf
 
Jekaterina Novikova - Machine Learning Methods in Detecting Alzheimer’s Disea...
Jekaterina Novikova - Machine Learning Methods in Detecting Alzheimer’s Disea...Jekaterina Novikova - Machine Learning Methods in Detecting Alzheimer’s Disea...
Jekaterina Novikova - Machine Learning Methods in Detecting Alzheimer’s Disea...
MLconf
 
Meghana Ravikumar - Optimized Image Classification on the Cheap
Meghana Ravikumar - Optimized Image Classification on the CheapMeghana Ravikumar - Optimized Image Classification on the Cheap
Meghana Ravikumar - Optimized Image Classification on the Cheap
MLconf
 
Noam Finkelstein - The Importance of Modeling Data Collection
Noam Finkelstein - The Importance of Modeling Data CollectionNoam Finkelstein - The Importance of Modeling Data Collection
Noam Finkelstein - The Importance of Modeling Data Collection
MLconf
 
June Andrews - The Uncanny Valley of ML
June Andrews - The Uncanny Valley of MLJune Andrews - The Uncanny Valley of ML
June Andrews - The Uncanny Valley of ML
MLconf
 
Sneha Rajana - Deep Learning Architectures for Semantic Relation Detection Tasks
Sneha Rajana - Deep Learning Architectures for Semantic Relation Detection TasksSneha Rajana - Deep Learning Architectures for Semantic Relation Detection Tasks
Sneha Rajana - Deep Learning Architectures for Semantic Relation Detection Tasks
MLconf
 
Anoop Deoras - Building an Incrementally Trained, Local Taste Aware, Global D...
Anoop Deoras - Building an Incrementally Trained, Local Taste Aware, Global D...Anoop Deoras - Building an Incrementally Trained, Local Taste Aware, Global D...
Anoop Deoras - Building an Incrementally Trained, Local Taste Aware, Global D...
MLconf
 
Vito Ostuni - The Voice: New Challenges in a Zero UI World
Vito Ostuni - The Voice: New Challenges in a Zero UI WorldVito Ostuni - The Voice: New Challenges in a Zero UI World
Vito Ostuni - The Voice: New Challenges in a Zero UI World
MLconf
 
Anna choromanska - Data-driven Challenges in AI: Scale, Information Selection...
Anna choromanska - Data-driven Challenges in AI: Scale, Information Selection...Anna choromanska - Data-driven Challenges in AI: Scale, Information Selection...
Anna choromanska - Data-driven Challenges in AI: Scale, Information Selection...
MLconf
 
Janani Kalyanam - Machine Learning to Detect Illegal Online Sales of Prescrip...
Janani Kalyanam - Machine Learning to Detect Illegal Online Sales of Prescrip...Janani Kalyanam - Machine Learning to Detect Illegal Online Sales of Prescrip...
Janani Kalyanam - Machine Learning to Detect Illegal Online Sales of Prescrip...
MLconf
 
Esperanza Lopez Aguilera - Using a Bayesian Neural Network in the Detection o...
Esperanza Lopez Aguilera - Using a Bayesian Neural Network in the Detection o...Esperanza Lopez Aguilera - Using a Bayesian Neural Network in the Detection o...
Esperanza Lopez Aguilera - Using a Bayesian Neural Network in the Detection o...
MLconf
 
Neel Sundaresan - Teaching a machine to code
Neel Sundaresan - Teaching a machine to codeNeel Sundaresan - Teaching a machine to code
Neel Sundaresan - Teaching a machine to code
MLconf
 
Rishabh Mehrotra - Recommendations in a Marketplace: Personalizing Explainabl...
Rishabh Mehrotra - Recommendations in a Marketplace: Personalizing Explainabl...Rishabh Mehrotra - Recommendations in a Marketplace: Personalizing Explainabl...
Rishabh Mehrotra - Recommendations in a Marketplace: Personalizing Explainabl...
MLconf
 
Soumith Chintala - Increasing the Impact of AI Through Better Software
Soumith Chintala - Increasing the Impact of AI Through Better SoftwareSoumith Chintala - Increasing the Impact of AI Through Better Software
Soumith Chintala - Increasing the Impact of AI Through Better Software
MLconf
 
Roy Lowrance - Predicting Bond Prices: Regime Changes
Roy Lowrance - Predicting Bond Prices: Regime ChangesRoy Lowrance - Predicting Bond Prices: Regime Changes
Roy Lowrance - Predicting Bond Prices: Regime Changes
MLconf
 

More from MLconf (20)

Jamila Smith-Loud - Understanding Human Impact: Social and Equity Assessments...
Jamila Smith-Loud - Understanding Human Impact: Social and Equity Assessments...Jamila Smith-Loud - Understanding Human Impact: Social and Equity Assessments...
Jamila Smith-Loud - Understanding Human Impact: Social and Equity Assessments...
 
Ted Willke - The Brain’s Guide to Dealing with Context in Language Understanding
Ted Willke - The Brain’s Guide to Dealing with Context in Language UnderstandingTed Willke - The Brain’s Guide to Dealing with Context in Language Understanding
Ted Willke - The Brain’s Guide to Dealing with Context in Language Understanding
 
Justin Armstrong - Applying Computer Vision to Reduce Contamination in the Re...
Justin Armstrong - Applying Computer Vision to Reduce Contamination in the Re...Justin Armstrong - Applying Computer Vision to Reduce Contamination in the Re...
Justin Armstrong - Applying Computer Vision to Reduce Contamination in the Re...
 
Igor Markov - Quantum Computing: a Treasure Hunt, not a Gold Rush
Igor Markov - Quantum Computing: a Treasure Hunt, not a Gold RushIgor Markov - Quantum Computing: a Treasure Hunt, not a Gold Rush
Igor Markov - Quantum Computing: a Treasure Hunt, not a Gold Rush
 
Josh Wills - Data Labeling as Religious Experience
Josh Wills - Data Labeling as Religious ExperienceJosh Wills - Data Labeling as Religious Experience
Josh Wills - Data Labeling as Religious Experience
 
Vinay Prabhu - Project GaitNet: Ushering in the ImageNet moment for human Gai...
Vinay Prabhu - Project GaitNet: Ushering in the ImageNet moment for human Gai...Vinay Prabhu - Project GaitNet: Ushering in the ImageNet moment for human Gai...
Vinay Prabhu - Project GaitNet: Ushering in the ImageNet moment for human Gai...
 
Jekaterina Novikova - Machine Learning Methods in Detecting Alzheimer’s Disea...
Jekaterina Novikova - Machine Learning Methods in Detecting Alzheimer’s Disea...Jekaterina Novikova - Machine Learning Methods in Detecting Alzheimer’s Disea...
Jekaterina Novikova - Machine Learning Methods in Detecting Alzheimer’s Disea...
 
Meghana Ravikumar - Optimized Image Classification on the Cheap
Meghana Ravikumar - Optimized Image Classification on the CheapMeghana Ravikumar - Optimized Image Classification on the Cheap
Meghana Ravikumar - Optimized Image Classification on the Cheap
 
Noam Finkelstein - The Importance of Modeling Data Collection
Noam Finkelstein - The Importance of Modeling Data CollectionNoam Finkelstein - The Importance of Modeling Data Collection
Noam Finkelstein - The Importance of Modeling Data Collection
 
June Andrews - The Uncanny Valley of ML
June Andrews - The Uncanny Valley of MLJune Andrews - The Uncanny Valley of ML
June Andrews - The Uncanny Valley of ML
 
Sneha Rajana - Deep Learning Architectures for Semantic Relation Detection Tasks
Sneha Rajana - Deep Learning Architectures for Semantic Relation Detection TasksSneha Rajana - Deep Learning Architectures for Semantic Relation Detection Tasks
Sneha Rajana - Deep Learning Architectures for Semantic Relation Detection Tasks
 
Anoop Deoras - Building an Incrementally Trained, Local Taste Aware, Global D...
Anoop Deoras - Building an Incrementally Trained, Local Taste Aware, Global D...Anoop Deoras - Building an Incrementally Trained, Local Taste Aware, Global D...
Anoop Deoras - Building an Incrementally Trained, Local Taste Aware, Global D...
 
Vito Ostuni - The Voice: New Challenges in a Zero UI World
Vito Ostuni - The Voice: New Challenges in a Zero UI WorldVito Ostuni - The Voice: New Challenges in a Zero UI World
Vito Ostuni - The Voice: New Challenges in a Zero UI World
 
Anna choromanska - Data-driven Challenges in AI: Scale, Information Selection...
Anna choromanska - Data-driven Challenges in AI: Scale, Information Selection...Anna choromanska - Data-driven Challenges in AI: Scale, Information Selection...
Anna choromanska - Data-driven Challenges in AI: Scale, Information Selection...
 
Janani Kalyanam - Machine Learning to Detect Illegal Online Sales of Prescrip...
Janani Kalyanam - Machine Learning to Detect Illegal Online Sales of Prescrip...Janani Kalyanam - Machine Learning to Detect Illegal Online Sales of Prescrip...
Janani Kalyanam - Machine Learning to Detect Illegal Online Sales of Prescrip...
 
Esperanza Lopez Aguilera - Using a Bayesian Neural Network in the Detection o...
Esperanza Lopez Aguilera - Using a Bayesian Neural Network in the Detection o...Esperanza Lopez Aguilera - Using a Bayesian Neural Network in the Detection o...
Esperanza Lopez Aguilera - Using a Bayesian Neural Network in the Detection o...
 
Neel Sundaresan - Teaching a machine to code
Neel Sundaresan - Teaching a machine to codeNeel Sundaresan - Teaching a machine to code
Neel Sundaresan - Teaching a machine to code
 
Rishabh Mehrotra - Recommendations in a Marketplace: Personalizing Explainabl...
Rishabh Mehrotra - Recommendations in a Marketplace: Personalizing Explainabl...Rishabh Mehrotra - Recommendations in a Marketplace: Personalizing Explainabl...
Rishabh Mehrotra - Recommendations in a Marketplace: Personalizing Explainabl...
 
Soumith Chintala - Increasing the Impact of AI Through Better Software
Soumith Chintala - Increasing the Impact of AI Through Better SoftwareSoumith Chintala - Increasing the Impact of AI Through Better Software
Soumith Chintala - Increasing the Impact of AI Through Better Software
 
Roy Lowrance - Predicting Bond Prices: Regime Changes
Roy Lowrance - Predicting Bond Prices: Regime ChangesRoy Lowrance - Predicting Bond Prices: Regime Changes
Roy Lowrance - Predicting Bond Prices: Regime Changes
 

Recently uploaded

GraphRAG is All You need? LLM & Knowledge Graph
GraphRAG is All You need? LLM & Knowledge GraphGraphRAG is All You need? LLM & Knowledge Graph
GraphRAG is All You need? LLM & Knowledge Graph
Guy Korland
 
Bits & Pixels using AI for Good.........
Bits & Pixels using AI for Good.........Bits & Pixels using AI for Good.........
Bits & Pixels using AI for Good.........
Alison B. Lowndes
 
Encryption in Microsoft 365 - ExpertsLive Netherlands 2024
Encryption in Microsoft 365 - ExpertsLive Netherlands 2024Encryption in Microsoft 365 - ExpertsLive Netherlands 2024
Encryption in Microsoft 365 - ExpertsLive Netherlands 2024
Albert Hoitingh
 
From Daily Decisions to Bottom Line: Connecting Product Work to Revenue by VP...
From Daily Decisions to Bottom Line: Connecting Product Work to Revenue by VP...From Daily Decisions to Bottom Line: Connecting Product Work to Revenue by VP...
From Daily Decisions to Bottom Line: Connecting Product Work to Revenue by VP...
Product School
 
Mission to Decommission: Importance of Decommissioning Products to Increase E...
Mission to Decommission: Importance of Decommissioning Products to Increase E...Mission to Decommission: Importance of Decommissioning Products to Increase E...
Mission to Decommission: Importance of Decommissioning Products to Increase E...
Product School
 
Key Trends Shaping the Future of Infrastructure.pdf
Key Trends Shaping the Future of Infrastructure.pdfKey Trends Shaping the Future of Infrastructure.pdf
Key Trends Shaping the Future of Infrastructure.pdf
Cheryl Hung
 
GDG Cloud Southlake #33: Boule & Rebala: Effective AppSec in SDLC using Deplo...
GDG Cloud Southlake #33: Boule & Rebala: Effective AppSec in SDLC using Deplo...GDG Cloud Southlake #33: Boule & Rebala: Effective AppSec in SDLC using Deplo...
GDG Cloud Southlake #33: Boule & Rebala: Effective AppSec in SDLC using Deplo...
James Anderson
 
Unsubscribed: Combat Subscription Fatigue With a Membership Mentality by Head...
Unsubscribed: Combat Subscription Fatigue With a Membership Mentality by Head...Unsubscribed: Combat Subscription Fatigue With a Membership Mentality by Head...
Unsubscribed: Combat Subscription Fatigue With a Membership Mentality by Head...
Product School
 
Software Delivery At the Speed of AI: Inflectra Invests In AI-Powered Quality
Software Delivery At the Speed of AI: Inflectra Invests In AI-Powered QualitySoftware Delivery At the Speed of AI: Inflectra Invests In AI-Powered Quality
Software Delivery At the Speed of AI: Inflectra Invests In AI-Powered Quality
Inflectra
 
FIDO Alliance Osaka Seminar: The WebAuthn API and Discoverable Credentials.pdf
FIDO Alliance Osaka Seminar: The WebAuthn API and Discoverable Credentials.pdfFIDO Alliance Osaka Seminar: The WebAuthn API and Discoverable Credentials.pdf
FIDO Alliance Osaka Seminar: The WebAuthn API and Discoverable Credentials.pdf
FIDO Alliance
 
De-mystifying Zero to One: Design Informed Techniques for Greenfield Innovati...
De-mystifying Zero to One: Design Informed Techniques for Greenfield Innovati...De-mystifying Zero to One: Design Informed Techniques for Greenfield Innovati...
De-mystifying Zero to One: Design Informed Techniques for Greenfield Innovati...
Product School
 
Accelerate your Kubernetes clusters with Varnish Caching
Accelerate your Kubernetes clusters with Varnish CachingAccelerate your Kubernetes clusters with Varnish Caching
Accelerate your Kubernetes clusters with Varnish Caching
Thijs Feryn
 
Elevating Tactical DDD Patterns Through Object Calisthenics
Elevating Tactical DDD Patterns Through Object CalisthenicsElevating Tactical DDD Patterns Through Object Calisthenics
Elevating Tactical DDD Patterns Through Object Calisthenics
Dorra BARTAGUIZ
 
Epistemic Interaction - tuning interfaces to provide information for AI support
Epistemic Interaction - tuning interfaces to provide information for AI supportEpistemic Interaction - tuning interfaces to provide information for AI support
Epistemic Interaction - tuning interfaces to provide information for AI support
Alan Dix
 
JMeter webinar - integration with InfluxDB and Grafana
JMeter webinar - integration with InfluxDB and GrafanaJMeter webinar - integration with InfluxDB and Grafana
JMeter webinar - integration with InfluxDB and Grafana
RTTS
 
LF Energy Webinar: Electrical Grid Modelling and Simulation Through PowSyBl -...
LF Energy Webinar: Electrical Grid Modelling and Simulation Through PowSyBl -...LF Energy Webinar: Electrical Grid Modelling and Simulation Through PowSyBl -...
LF Energy Webinar: Electrical Grid Modelling and Simulation Through PowSyBl -...
DanBrown980551
 
UiPath Test Automation using UiPath Test Suite series, part 3
UiPath Test Automation using UiPath Test Suite series, part 3UiPath Test Automation using UiPath Test Suite series, part 3
UiPath Test Automation using UiPath Test Suite series, part 3
DianaGray10
 
When stars align: studies in data quality, knowledge graphs, and machine lear...
When stars align: studies in data quality, knowledge graphs, and machine lear...When stars align: studies in data quality, knowledge graphs, and machine lear...
When stars align: studies in data quality, knowledge graphs, and machine lear...
Elena Simperl
 
PCI PIN Basics Webinar from the Controlcase Team
PCI PIN Basics Webinar from the Controlcase TeamPCI PIN Basics Webinar from the Controlcase Team
PCI PIN Basics Webinar from the Controlcase Team
ControlCase
 
The Art of the Pitch: WordPress Relationships and Sales
The Art of the Pitch: WordPress Relationships and SalesThe Art of the Pitch: WordPress Relationships and Sales
The Art of the Pitch: WordPress Relationships and Sales
Laura Byrne
 

Recently uploaded (20)

GraphRAG is All You need? LLM & Knowledge Graph
GraphRAG is All You need? LLM & Knowledge GraphGraphRAG is All You need? LLM & Knowledge Graph
GraphRAG is All You need? LLM & Knowledge Graph
 
Bits & Pixels using AI for Good.........
Bits & Pixels using AI for Good.........Bits & Pixels using AI for Good.........
Bits & Pixels using AI for Good.........
 
Encryption in Microsoft 365 - ExpertsLive Netherlands 2024
Encryption in Microsoft 365 - ExpertsLive Netherlands 2024Encryption in Microsoft 365 - ExpertsLive Netherlands 2024
Encryption in Microsoft 365 - ExpertsLive Netherlands 2024
 
From Daily Decisions to Bottom Line: Connecting Product Work to Revenue by VP...
From Daily Decisions to Bottom Line: Connecting Product Work to Revenue by VP...From Daily Decisions to Bottom Line: Connecting Product Work to Revenue by VP...
From Daily Decisions to Bottom Line: Connecting Product Work to Revenue by VP...
 
Mission to Decommission: Importance of Decommissioning Products to Increase E...
Mission to Decommission: Importance of Decommissioning Products to Increase E...Mission to Decommission: Importance of Decommissioning Products to Increase E...
Mission to Decommission: Importance of Decommissioning Products to Increase E...
 
Key Trends Shaping the Future of Infrastructure.pdf
Key Trends Shaping the Future of Infrastructure.pdfKey Trends Shaping the Future of Infrastructure.pdf
Key Trends Shaping the Future of Infrastructure.pdf
 
GDG Cloud Southlake #33: Boule & Rebala: Effective AppSec in SDLC using Deplo...
GDG Cloud Southlake #33: Boule & Rebala: Effective AppSec in SDLC using Deplo...GDG Cloud Southlake #33: Boule & Rebala: Effective AppSec in SDLC using Deplo...
GDG Cloud Southlake #33: Boule & Rebala: Effective AppSec in SDLC using Deplo...
 
Unsubscribed: Combat Subscription Fatigue With a Membership Mentality by Head...
Unsubscribed: Combat Subscription Fatigue With a Membership Mentality by Head...Unsubscribed: Combat Subscription Fatigue With a Membership Mentality by Head...
Unsubscribed: Combat Subscription Fatigue With a Membership Mentality by Head...
 
Software Delivery At the Speed of AI: Inflectra Invests In AI-Powered Quality
Software Delivery At the Speed of AI: Inflectra Invests In AI-Powered QualitySoftware Delivery At the Speed of AI: Inflectra Invests In AI-Powered Quality
Software Delivery At the Speed of AI: Inflectra Invests In AI-Powered Quality
 
FIDO Alliance Osaka Seminar: The WebAuthn API and Discoverable Credentials.pdf
FIDO Alliance Osaka Seminar: The WebAuthn API and Discoverable Credentials.pdfFIDO Alliance Osaka Seminar: The WebAuthn API and Discoverable Credentials.pdf
FIDO Alliance Osaka Seminar: The WebAuthn API and Discoverable Credentials.pdf
 
De-mystifying Zero to One: Design Informed Techniques for Greenfield Innovati...
De-mystifying Zero to One: Design Informed Techniques for Greenfield Innovati...De-mystifying Zero to One: Design Informed Techniques for Greenfield Innovati...
De-mystifying Zero to One: Design Informed Techniques for Greenfield Innovati...
 
Accelerate your Kubernetes clusters with Varnish Caching
Accelerate your Kubernetes clusters with Varnish CachingAccelerate your Kubernetes clusters with Varnish Caching
Accelerate your Kubernetes clusters with Varnish Caching
 
Elevating Tactical DDD Patterns Through Object Calisthenics
Elevating Tactical DDD Patterns Through Object CalisthenicsElevating Tactical DDD Patterns Through Object Calisthenics
Elevating Tactical DDD Patterns Through Object Calisthenics
 
Epistemic Interaction - tuning interfaces to provide information for AI support
Epistemic Interaction - tuning interfaces to provide information for AI supportEpistemic Interaction - tuning interfaces to provide information for AI support
Epistemic Interaction - tuning interfaces to provide information for AI support
 
JMeter webinar - integration with InfluxDB and Grafana
JMeter webinar - integration with InfluxDB and GrafanaJMeter webinar - integration with InfluxDB and Grafana
JMeter webinar - integration with InfluxDB and Grafana
 
LF Energy Webinar: Electrical Grid Modelling and Simulation Through PowSyBl -...
LF Energy Webinar: Electrical Grid Modelling and Simulation Through PowSyBl -...LF Energy Webinar: Electrical Grid Modelling and Simulation Through PowSyBl -...
LF Energy Webinar: Electrical Grid Modelling and Simulation Through PowSyBl -...
 
UiPath Test Automation using UiPath Test Suite series, part 3
UiPath Test Automation using UiPath Test Suite series, part 3UiPath Test Automation using UiPath Test Suite series, part 3
UiPath Test Automation using UiPath Test Suite series, part 3
 
When stars align: studies in data quality, knowledge graphs, and machine lear...
When stars align: studies in data quality, knowledge graphs, and machine lear...When stars align: studies in data quality, knowledge graphs, and machine lear...
When stars align: studies in data quality, knowledge graphs, and machine lear...
 
PCI PIN Basics Webinar from the Controlcase Team
PCI PIN Basics Webinar from the Controlcase TeamPCI PIN Basics Webinar from the Controlcase Team
PCI PIN Basics Webinar from the Controlcase Team
 
The Art of the Pitch: WordPress Relationships and Sales
The Art of the Pitch: WordPress Relationships and SalesThe Art of the Pitch: WordPress Relationships and Sales
The Art of the Pitch: WordPress Relationships and Sales
 

Interpretability beyond feature attribution quantitative testing with concept activation vectors (tcav)

  • 1. Interpretability beyond feature attribution: Testing with Concept Activation Vectors TCAV Been Kim Presenting work with a lot of awesome people inside and outside of Google: Marten Wattenberg, Julius Adebayo, Justin Gilmer, Carrie Cai, James Wexler, Fernanda Viegas, Ian Goodfellow, Mortiz Hardt, Rory Sayres
  • 2. My goal interpretability !2 To use machine learning responsibly we need to ensure that 1. our values are aligned 2. our knowledge is reflected for everyone. http://blogs.teradata.com/ Machine Learning Models Human
  • 3. My goal interpretability !3 To use machine learning responsibly we need to ensure that 1. our values are aligned 2. our knowledge is reflected for everyone. http://blogs.teradata.com/ Machine Learning Models Human
  • 4. ingredients for interpretability methods. !4 ? Some quality function
  • 9. Class1 Class0 Human What’s ML? If I were you, I would train a neural network. newbie expert
  • 10. Class1 Class0 Task • Local vs. global • Simple explanations vs. more complex but more accurate explanations • Low or high stake domains
  • 11. Interpretability methods Post-training explanations Building inherently interpretable models Explaining data
  • 13. Agenda 1. Revisiting existing methods: Saliency maps 2. Making explanations using the way humans think: Testing with concept activation vectors (TCAV) Post-training explanations
  • 14. Agenda Sanity Checks for Saliency Maps Joint work with Adebayo, Gilmer, Goodfellow, Hardt, [NIPS 18] 1. Revisiting existing methods: Saliency maps 2. Making explanations using the way humans think: Testing with concept activation vectors (TCAV) Post-training explanations
  • 15. Problem: Post-training explanation !15 cash-machine-ness A trained machine learning model (e.g., neural network) Why was this a cash machine?
  • 16. One of the most popular interpretability methods for images: Saliency maps !16 SmoothGrad [Smilkov, Thorat, K., Viégas, Wattenberg ’17] Integrated gradient [Sundararajan, Taly, Yan ’17] picture credit: @sayres Caaaaan do! We’ve got saliency maps to measure importance of each pixel! a logit pixel i,j
  • 17. One of the most popular interpretability methods for images: Saliency maps !17 SmoothGrad [Smilkov, Thorat, K., Viégas, Wattenberg ’17] Integrated gradient [Sundararajan, Taly, Yan ’17] widely used for images local undestandingNN humans’ subjective judgement picture credit: @sayres
  • 18. One of the most popular interpretability methods for images: Saliency maps !18 Sanity check: If I change M a lot, will human perceive that E has changed a lot? SmoothGrad [Smilkov, Thorat, K., Viégas, Wattenberg ’17] Integrated gradient [Sundararajan, Taly, Yan ’17] widely used for images local undestandingNN humans’ subjective judgement picture credit: @sayres
  • 19. Some confusing behaviors of saliency maps. Saliency map Sanity Checks for Saliency Maps Joint work with Adebayo, Gilmer, Goodfellow, Hardt, [NIPS 18]
  • 20. Some confusing behaviors of saliency maps. Randomized weights! Network now makes garbage prediction. Saliency map Sanity Checks for Saliency Maps Joint work with Adebayo, Gilmer, Goodfellow, Hardt, [NIPS 18]
  • 21. Some confusing behaviors of saliency maps. Saliency map Randomized weights! Network now makes garbage prediction. !!!!!???!? Sanity Checks for Saliency Maps Joint work with Adebayo, Gilmer, Goodfellow, Hardt, [NIPS 18]
  • 22. Some saliency maps look similar when we randomize the network (= making the network completely useless) Before After Guided Backprop Integrated Gradient Sanity Checks for Saliency Maps Joint work with Adebayo, Gilmer, Goodfellow, Hardt, [NIPS 18]
  • 23. • Potential human confirmation bias: Just because it “makes sense” to humans, doesn’t mean they reflect evidence for the prediction. • Our discovery is consistent with other findings. [Nie, Zhang, Patel ’18] [Ulyanov, Vedaldi, Lempitsky ’18] • Some of these methods have been shown to be useful for humans. Why? More studies needed. What can we learn from this? Sanity Checks for Saliency Maps Joint work with Adebayo, Gilmer, Goodfellow, Hardt, [NIPS 18]
  • 24. !24 local undestanding lay person? human’s subjective judgement What can we do better? Creating a wishlist. Using input features as a language
  • 26. Agenda TCAV [ICML’18] Joint work with Wattenberg, Gilmer, Cai, Wexler, Viegas, Sayres 1. Revisiting existing methods: Saliency maps 2. Making explanations using the way humans think: Testing with concept activation vectors (TCAV) Post-training explanations
  • 27. Problem: Post-training explanation !27 cash-machine-ness A trained machine learning model (e.g., neural network) Why was this a cash machine? TCAV [ICML’18] Joint work with Wattenberg, Gilmer, Cai, Wexler, Viegas, Sayres
  • 28. prediction: Cash machine https://pair-code.github.io/saliency/ SmoothGrad [Smilkov, Thorat, K., Viégas, Wattenberg ’17] Common solution: Saliency map !28 Let’s use this to help us think about what what we really want to ask.
  • 29. prediction: Cash machine https://pair-code.github.io/saliency/ SmoothGrad [Smilkov, Thorat, K., Viégas, Wattenberg ’17] What we really want to ask… !29 Were there more pixels on the cash machine than on the person? Did the ‘human’ concept matter? Did the ‘wheels’ concept matter?
  • 30. prediction: Cash machine https://pair-code.github.io/saliency/ SmoothGrad [Smilkov, Thorat, K., Viégas, Wattenberg ’17] What we really want to ask… !30 Were there more pixels on the cash machine than on the person? Which concept mattered more? Is this true for all other cash machine predictions? Did the ‘human’ concept matter? Did the ‘wheels’ concept matter?
  • 31. prediction: Cash machine https://pair-code.github.io/saliency/ SmoothGrad [Smilkov, Thorat, K., Viégas, Wattenberg ’17] What we really want to ask… !31 Oh no! I can’t express these concepts as pixels!! They weren’t my input features either! Were there more pixels on the cash machine than on the person? Which concept mattered more? Is this true for all other cash machine predictions? Did the ‘human’ concept matter? Did the ‘wheels’ concept matter?
  • 32. prediction: Cash machine https://pair-code.github.io/saliency/ SmoothGrad [Smilkov, Thorat, K., Viégas, Wattenberg ’17] What we really want to ask… !32 Were there more pixels on the cash machine than on the person? Which concept mattered more? Is this true for all other cash machine predictions? Wouldn’t it be great if we can quantitatively measure how important any of these user-chosen concepts are? Did the ‘human’ concept matter? Did the ‘wheels’ concept matter?
  • 33. Quantitative explanation: how much a concept (e.g., gender, race) was important for a prediction in a trained model. …even if the concept was not part of the training. Goal of TCAV: Testing with Concept Activation Vectors !33 ICML 2018
  • 34. Goal of TCAV: Testing with Concept Activation Vectors !34 Doctor-ness A trained machine learning model (e.g., neural network) vactruth.com healthcommunitiesproviderservices
  • 35. Goal of TCAV: Testing with Concept Activation Vectors !35 Doctor-ness A trained machine learning model (e.g., neural network) Was gender concept important to this doctor image classifier? vactruth.com healthcommunitiesproviderservices
  • 36. Goal of TCAV: Testing with Concept Activation Vectors !36 Doctor-ness TCAV score for womennot women Doctor A trained machine learning model (e.g., neural network) vactruth.com healthcommunitiesproviderservices Was gender concept important to this doctor image classifier?
  • 37. Goal of TCAV: Testing with Concept Activation Vectors !37 Doctor-ness TCAV score for womennot women Doctor A trained machine learning model (e.g., neural network) vactruth.com healthcommunitiesproviderservices Was gender concept important to this doctor image classifier? TCAV provides quantitative importance of a concept if and only if your network learned about it.
  • 38. Goal of TCAV: Testing with Concept Activation Vectors !38 zebra-ness A trained machine learning model (e.g., neural network) Was striped concept important to this zebra image classifier? TCAV score for not stripedstriped Zebra TCAV provides quantitative importance of a concept if and only if your network learned about it.
  • 39. TCAV TCAV: Testing with Concept Activation Vectors !39 zebra-ness A trained machine learning model (e.g., neural network) Was striped concept important to this zebra image classifier? 1. Learning CAVs 1. How to define concepts?
  • 40. Defining concept activation vector (CAV) Inputs: !40 Random images Examples of concepts A trained network under investigation and Internal tensors
  • 41. !41 Inputs: Train a linear classifier to separate activations. CAV ( ) is the vector orthogonal to the decision boundary. [Smilkov ’17, Bolukbasi ’16 , Schmidt ’15] Defining concept activation vector (CAV)
  • 42. TCAV TCAV: Testing with Concept Activation Vectors !42 zebra-ness A trained machine learning model (e.g., neural network) Was striped concept important to this zebra image classifier? 1. Learning CAVs 2. Getting TCAV score 2. How are the CAVs useful to get explanations?
  • 43. TCAV core idea: Derivative with CAV to get prediction sensitivity !43 TCAV score Directional derivative with CAV
  • 44. striped CAV TCAV score Directional derivative with CAV TCAV core idea: Derivative with CAV to get prediction sensitivity !44
  • 45. TCAV TCAV: Testing with Concept Activation Vectors !45 zebra-ness A trained machine learning model (e.g., neural network) Was striped concept important to this zebra image classifier? 1. Learning CAVs 2. Getting TCAV score
  • 46. TCAV TCAV: Testing with Concept Activation Vectors !46 zebra-ness A trained machine learning model (e.g., neural network) Was striped concept important to this zebra image classifier? 1. Learning CAVs 2. Getting TCAV score 3. CAV validation Qualitative Quantitative
  • 47. Quantitative validation: Guarding against spurious CAV Did my CAVs returned high sensitivity by chance? !47
  • 48. Learn many stripes CAVs using different sets of random images Quantitative validation: Guarding against spurious CAV !48
  • 51. Check the distribution of is statistically different from random using t-test TCAV score random …… Zebra Quantitative validation: Guarding against spurious CAV !51 *
  • 52. Recap TCAV: Testing with Concept Activation Vectors !52 1. Learning CAVs 2. Getting TCAV score 3. CAV validation Qualitative Quantitative TCAV provides quantitative importance of a concept if and only if your network learned about it. Even if your training data wasn’t tagged with the concept Even if your input feature did not include the concept
  • 53. Results 1. Sanity check experiment 2. Biases in Inception V3 and GoogleNet 3. Domain expert confirmation from Diabetic Retinopathy !53
  • 54. Results 1. Sanity check experiment 2. Biases from Inception V3 and GoogleNet 3. Domain expert confirmation from Diabetic Retinopathy !54
  • 55. Sanity check experiment !55 If we know the ground truth (important concepts), will TCAV match?
  • 56. Sanity check experiment setup !56 An image + Potentially noisy Caption
  • 57. Sanity check experiment setup !57 An image + Potentially noisy Caption image concept models can use either image or caption concept for classification. caption concept
  • 58. Sanity check experiment setup !58 Caption noise level in training set An image + Potentially noisy Caption image concept models can use either image or caption concept for classification. caption concept
  • 59. Sanity check experiment setup !59 Test accuracy with no caption image = Importance of image concept Caption noise level in training set image concept caption concept models can use either image or caption concept for classification.
  • 60. Sanity check experiment !60 Caption noise level in training set Caption noise level in training set Test accuracy with no caption image
  • 61. !61 Cool, cool. Can saliency maps do this too?
  • 62. Can saliency maps communicate the same information? !62 Ground truth Image concept Image concept Image concept Image concept Image with caption
  • 63. Human subject experiment: Can saliency maps communicate the same information? • 50 turkers are • asked to judge importance of image vs. c. ept given saliency maps. • asked to indicate their confidence • shown 3 classes (cab, zebra, cucumber) x 2 saliency maps for one model !63 image caption
  • 64. !64 • Random chance: 50% • Human performance with saliency map: 52% • Humans can’t agree: more than 50% no significant consensus • Humans are very confident even when they are wrong. Human subject experiment: Can saliency maps communicate the same information?
  • 65. Human subject experiment: Can saliency maps communicate the same information? • Random chance: 50% • Human performance with saliency map: 52% • Humans can’t agree: more than 50% no significant consensus • Humans are very confident even when they are wrong. !65
  • 66. Results 1. Sanity check experiment 2. Biases from Inception V3 and GoogleNet 3. Domain expert confirmation from Diabetic Retinopathy !66
  • 67. TCAV in Two widely used image prediction models !67
  • 68. TCAV in Two widely used image prediction models !68 Geographical bias! http://www.abc.net.au
  • 69. TCAV in Two widely used image prediction models !69 Quantitative confirmation to previously qualitative findings [Stock & Cisse, 2017] Geographical bias?
  • 70. TCAV in Two widely used image prediction models !70 Quantitative confirmation to previously qualitative findings [Stock & Cisse, 2017] Geographical bias? Goal of interpretability: To use machine learning responsibly we need to ensure that 1. our values are aligned 2. our knowledge is reflected
  • 71. Results 1. Sanity check experiment 2. Biases Inception V3 and GoogleNet 3. Domain expert confirmation from Diabetic Retinopathy !71
  • 72. Diabetic Retinopathy • Treatable but sight-threatening conditions • Have model to with accurate prediction of DR (85%) [Krause et al., 2017] !72 Concepts the ML model uses Vs Diagnostic Concepts human doctors use
  • 73. Collect human doctor’s knowledge !73 PRP PRH/VH NV/FP VB MA HMA DR level 4 DR level 1 Concepts belong to this level Concepts do not belong to this level
  • 74. TCAV for Diabetic Retinopathy !74 PRP PRH/VH NV/FP VB Green: domain expert’s label on concepts belong to the level Red: domain expert’s label on concepts does not belong to the level Prediction class DR level 4 Prediction accuracy High Example TCAV scores TCAV shows the model is consistent with doctor’s knowledge when model is accurate
  • 75. PRP PRH/VH NV/FP VB Green: domain expert’s label on concepts belong to the level Red: domain expert’s label on concepts does not belong to the level Prediction class DR level 4 Prediction accuracy High Example TCAV scores TCAV shows the model is consistent with doctor’s knowledge when model is accurate TCAV shows the model is inconsistent with doctor’s knowledge for classes when model is less accurate DR level 1 Med TCAV for Diabetic Retinopathy !75 MA HMA
  • 76. PRP PRH/VH NV/FP VB Green: domain expert’s label on concepts belong to the level Red: domain expert’s label on concepts does not belong to the level Prediction class DR level 4 Prediction accuracy High Example TCAV scores TCAV shows the model is consistent with doctor’s knowledge when model is accurate Level 1 was often confused to level 2. DR level 1 Low TCAV shows the model is inconsistent with doctor’s knowledge for classes when model is less accurate TCAV for Diabetic Retinopathy !76 MA HMA Goal of interpretability: To use machine learning responsibly we need to ensure that 1. our values are aligned 2. our knowledge is reflected
  • 77. Summary: Testing with Concept Activation Vectors !77 stripes concept (score: 0.9) was important to zebra class for this trained network. PRP PRH/VH NV/FP VB Our values Our knowledge TCAV provides quantitative importance of a concept if and only if your network learned about it. Joint work with Wattenberg, Gilmer, Cai, Wexler, Viegas, Sayres ICML 2018