SlideShare une entreprise Scribd logo
1  sur  21
Télécharger pour lire hors ligne
Boosting DL Concept Learners
Nicola FANIZZI, Giuseppe RIZZO and Claudia d'AMATO
LACAM — Dipartimento di Informatica & CILA
Università degli studi di Bari Aldo Moro
ESWC 2019 — June 4th, 2019
Outline
Introduction
The Framework:
Training
Strong learner
Weak learner
Test
Empirical Evaluation
Conclusions and Future Work
Introduction
Context & Motivations
Considering knowledge bases in the Web of Data
Goal: Inducing a classification function (predictive model) for a
target concept to assess the membership of (unseen) individuals
Supervised approach: each (training) example labeled as
positive/negative/uncertain
Possible Applications: approximate classification, semi-automatic KB
completion (assertional part)
Context & Motivations
Several approaches:
statistical approaches ⟶ prediction
concept learners ⟶ prediction + intensional definition
Poor model generalization (limited predictiveness)⚠
Context & Motivations
Weak Model 1
Strong Model
Weak Model 2
⋮ ⋮
⋮ ⋮
Weak Model n
Strong model more predictive than weak models
Ensemble Classification (boosting, bagging,...)ⓘ
Context & Motivations
Contribution
A framework for confidence-rated boosting for DL concept
learners
pool of weak models as concepts with the related score
each model induced to cover hardest examples (leading to
misclassifications)
Example
Learn a classification function for the individuals w.r.t. the target
concept .
Training set :
(instances of with a known child),
(instances of )
Ensemble Model:
Father
K = {
{
Man ⊑ Person, Woman ⊑ Person, Artist ⊑ Person, Man ⊑ ¬Woman} ∪
Man(a), Man(b), Man(c), Woman(d), Woman(f), Artist(e), Robot(g),
hasChild(a, d), hasChild(b, e)}
T
Ps = {a, b} Man
Ns = {d, f} Woman
Us = {c, e, g}
{Person [0.4], ¬Woman [0.8], Person [0.5], ∃hasChild. ⊤ [0.9]}
The Boosting Framework
Proposed Solution
Given and a set of training examples
Weight assigned to example (init. )
Ensemble of weak models (concepts) induced by a weak
learner ( DLF):
Confidence computed per weak model
Weights updated according to the coverage of the hypothesis
concept
decreased for examples covered by
increased for examples uncovered by
K T
wi ai = 1/|T|w
(0)
i
T Ct
ct
Ct
Ct
Coverage: Definition
An example covered by when
i.e., is correctly classified
x Ct
⎧
⎩
⎨
K ⊨ (x) if x ∈ PsCt
K ⊨ ¬ (x) if x ∈ NsCt
K ⊭ (x) ∧ K ⊭ ¬ (x) if x ∈ UsCt Ct
x
wDLF: Training
Given , weights associated to the examples, and a partial
description ( )
specialized via ref. operator ⟶ a set of concepts
score: difference between sum of the weights of the covered
ex.s and uncovered ex.s
heuristic: maximum difference between the scores of and
(gain)
T w
(t)
Ct = ⊤Ct
Ct ρ ⊑C
′
t
Ct
−∑
i| coveredai
wi
− −−−−−−−−
√
∑
i| uncoveredai
wi
− −−−−−−−−−−
√
C
′
t
Ct
Weight Updates
After is returned, weights of the examples updated:
for each example the new weight will be
where
with depending on the (ratio between the)sum of the weights of
covered and the uncovered (optimal value to minimize the training error)
update
Multiplied by for uncovered examples
Divided by for covered examples
Ct
ai w
(t)
i
= exp(− )w
(t)
i
w
(t)
i
ct
= − ⋅ (−1ct α^t )
I( , ( ))li ht ei
α^
exp ct
exp ct
Prediction
Given the model and a query individual
Prediction of single weak model :
if
if
otherwise
Final prediction: label according to weighted majority vote
sums of confidence values for the three labels
{ [ ]}Ct ct
T
t=1
a
Ct
+1 K ⊨ (a)Ct
−1 K ⊨ ¬ (a)Ct
0
=l
∗
argmax
l∈{+1,−1,0}
∑
t∈T
p( (x)=l)h
t
ct
Experimental Evaluation
Evaluation: Settings
5 KBs for producing 15 artificial problems/target concepts (per KB)
and related training & test sets
Parameters:
Ensemble size:
refinements generated via wDLF:
Baseline: wDLF
Other systems: C and DL-F
Boostrap procedure: 30 replications per learning problems/target
Evaluation indices:
match: exact predictions
commission: opposite predictions
omission: non definite predictions [for -label exs]
induction: definite predictions [for -label exs]
10
30
±1
0
Evaluation: Outcomes
Evaluation: Examples of Induced Models
Conclusions & Future Work
Conclusions
A boosted ensemble classifier
how to derive a weak learner
good performance for complex target concepts
Future Work
other (weak) learners exploiting ref.ops
dynamic ensemble dimension (validation set)
approximate reasoners for efficiency
formatted by Markdeep 1.04
✒

Contenu connexe

Similaire à Boosting dl concept learners

Data.Mining.C.6(II).classification and prediction
Data.Mining.C.6(II).classification and predictionData.Mining.C.6(II).classification and prediction
Data.Mining.C.6(II).classification and predictionMargaret Wang
 
Rd1 r17a19 datawarehousing and mining_cap617t_cap617
Rd1 r17a19 datawarehousing and mining_cap617t_cap617Rd1 r17a19 datawarehousing and mining_cap617t_cap617
Rd1 r17a19 datawarehousing and mining_cap617t_cap617Ravi Kumar
 
20070702 Text Categorization
20070702 Text Categorization20070702 Text Categorization
20070702 Text Categorizationmidi
 
Artificial Intelligence.pptx
Artificial Intelligence.pptxArtificial Intelligence.pptx
Artificial Intelligence.pptxKaviya452563
 
Adaptive Multilevel Clustering Model for the Prediction of Academic Risk
Adaptive Multilevel Clustering Model for the Prediction of Academic RiskAdaptive Multilevel Clustering Model for the Prediction of Academic Risk
Adaptive Multilevel Clustering Model for the Prediction of Academic RiskXavier Ochoa
 
Lecture 09(introduction to machine learning)
Lecture 09(introduction to machine learning)Lecture 09(introduction to machine learning)
Lecture 09(introduction to machine learning)Jeet Das
 
Introduction to Machine Learning Lectures
Introduction to Machine Learning LecturesIntroduction to Machine Learning Lectures
Introduction to Machine Learning Lecturesssuserfece35
 
slides
slidesslides
slidesbutest
 
slides
slidesslides
slidesbutest
 
Learning On The Border:Active Learning in Imbalanced classification Data
Learning On The Border:Active Learning in Imbalanced classification DataLearning On The Border:Active Learning in Imbalanced classification Data
Learning On The Border:Active Learning in Imbalanced classification Data萍華 楊
 
3_learning.ppt
3_learning.ppt3_learning.ppt
3_learning.pptbutest
 
Lecture 6: Ensemble Methods
Lecture 6: Ensemble Methods Lecture 6: Ensemble Methods
Lecture 6: Ensemble Methods Marina Santini
 
Supervised Machine learning Algorithm.pptx
Supervised Machine learning Algorithm.pptxSupervised Machine learning Algorithm.pptx
Supervised Machine learning Algorithm.pptxKing Khalid University
 
supervised-learning.pptx
supervised-learning.pptxsupervised-learning.pptx
supervised-learning.pptxGandhiMathy6
 
Meta Pseudo Label - InsideAIML
Meta Pseudo Label - InsideAIMLMeta Pseudo Label - InsideAIML
Meta Pseudo Label - InsideAIMLVijaySharma802
 

Similaire à Boosting dl concept learners (20)

Data.Mining.C.6(II).classification and prediction
Data.Mining.C.6(II).classification and predictionData.Mining.C.6(II).classification and prediction
Data.Mining.C.6(II).classification and prediction
 
.ppt
.ppt.ppt
.ppt
 
Rd1 r17a19 datawarehousing and mining_cap617t_cap617
Rd1 r17a19 datawarehousing and mining_cap617t_cap617Rd1 r17a19 datawarehousing and mining_cap617t_cap617
Rd1 r17a19 datawarehousing and mining_cap617t_cap617
 
boosting algorithm
boosting algorithmboosting algorithm
boosting algorithm
 
20070702 Text Categorization
20070702 Text Categorization20070702 Text Categorization
20070702 Text Categorization
 
18.1 combining models
18.1 combining models18.1 combining models
18.1 combining models
 
Unit-2.ppt
Unit-2.pptUnit-2.ppt
Unit-2.ppt
 
Artificial Intelligence.pptx
Artificial Intelligence.pptxArtificial Intelligence.pptx
Artificial Intelligence.pptx
 
Adaptive Multilevel Clustering Model for the Prediction of Academic Risk
Adaptive Multilevel Clustering Model for the Prediction of Academic RiskAdaptive Multilevel Clustering Model for the Prediction of Academic Risk
Adaptive Multilevel Clustering Model for the Prediction of Academic Risk
 
Lecture 09(introduction to machine learning)
Lecture 09(introduction to machine learning)Lecture 09(introduction to machine learning)
Lecture 09(introduction to machine learning)
 
Introduction to Machine Learning Lectures
Introduction to Machine Learning LecturesIntroduction to Machine Learning Lectures
Introduction to Machine Learning Lectures
 
slides
slidesslides
slides
 
slides
slidesslides
slides
 
Learning On The Border:Active Learning in Imbalanced classification Data
Learning On The Border:Active Learning in Imbalanced classification DataLearning On The Border:Active Learning in Imbalanced classification Data
Learning On The Border:Active Learning in Imbalanced classification Data
 
3_learning.ppt
3_learning.ppt3_learning.ppt
3_learning.ppt
 
Lecture 6: Ensemble Methods
Lecture 6: Ensemble Methods Lecture 6: Ensemble Methods
Lecture 6: Ensemble Methods
 
Clustering
ClusteringClustering
Clustering
 
Supervised Machine learning Algorithm.pptx
Supervised Machine learning Algorithm.pptxSupervised Machine learning Algorithm.pptx
Supervised Machine learning Algorithm.pptx
 
supervised-learning.pptx
supervised-learning.pptxsupervised-learning.pptx
supervised-learning.pptx
 
Meta Pseudo Label - InsideAIML
Meta Pseudo Label - InsideAIMLMeta Pseudo Label - InsideAIML
Meta Pseudo Label - InsideAIML
 

Plus de Giuseppe Rizzo

DL-Foil:Class Expression Learning Revisited
DL-Foil:Class Expression Learning RevisitedDL-Foil:Class Expression Learning Revisited
DL-Foil:Class Expression Learning RevisitedGiuseppe Rizzo
 
A framework for Tackling myopia in concept learning on the Web of Data
A framework for Tackling myopia in concept learning on the Web of DataA framework for Tackling myopia in concept learning on the Web of Data
A framework for Tackling myopia in concept learning on the Web of DataGiuseppe Rizzo
 
Terminological cluster trees for Disjointness Axiom Discovery
Terminological cluster trees for Disjointness Axiom DiscoveryTerminological cluster trees for Disjointness Axiom Discovery
Terminological cluster trees for Disjointness Axiom DiscoveryGiuseppe Rizzo
 
Approximating Numeric Role Fillers via Predictive Clustering Trees for Know...
Approximating Numeric Role Fillers via Predictive Clustering Trees  for  Know...Approximating Numeric Role Fillers via Predictive Clustering Trees  for  Know...
Approximating Numeric Role Fillers via Predictive Clustering Trees for Know...Giuseppe Rizzo
 
Inducing Predictive Clustering Trees for Datatype properties Values
Inducing Predictive Clustering Trees for Datatype properties ValuesInducing Predictive Clustering Trees for Datatype properties Values
Inducing Predictive Clustering Trees for Datatype properties ValuesGiuseppe Rizzo
 
On the Effectiveness of Evidence-based Terminological Decision Trees
On the Effectiveness of Evidence-based Terminological Decision TreesOn the Effectiveness of Evidence-based Terminological Decision Trees
On the Effectiveness of Evidence-based Terminological Decision TreesGiuseppe Rizzo
 
Inductive Classification through Evidence-based Models and Their Ensemble
Inductive Classification through Evidence-based Models and Their EnsembleInductive Classification through Evidence-based Models and Their Ensemble
Inductive Classification through Evidence-based Models and Their EnsembleGiuseppe Rizzo
 
Tackling the Class Imbalance Learning Problem in Semantic Web Knowledge bases
Tackling the Class Imbalance Learning Problem in Semantic Web Knowledge basesTackling the Class Imbalance Learning Problem in Semantic Web Knowledge bases
Tackling the Class Imbalance Learning Problem in Semantic Web Knowledge basesGiuseppe Rizzo
 
Towards Evidence Terminological Decision Tree
Towards Evidence Terminological Decision TreeTowards Evidence Terminological Decision Tree
Towards Evidence Terminological Decision TreeGiuseppe Rizzo
 

Plus de Giuseppe Rizzo (9)

DL-Foil:Class Expression Learning Revisited
DL-Foil:Class Expression Learning RevisitedDL-Foil:Class Expression Learning Revisited
DL-Foil:Class Expression Learning Revisited
 
A framework for Tackling myopia in concept learning on the Web of Data
A framework for Tackling myopia in concept learning on the Web of DataA framework for Tackling myopia in concept learning on the Web of Data
A framework for Tackling myopia in concept learning on the Web of Data
 
Terminological cluster trees for Disjointness Axiom Discovery
Terminological cluster trees for Disjointness Axiom DiscoveryTerminological cluster trees for Disjointness Axiom Discovery
Terminological cluster trees for Disjointness Axiom Discovery
 
Approximating Numeric Role Fillers via Predictive Clustering Trees for Know...
Approximating Numeric Role Fillers via Predictive Clustering Trees  for  Know...Approximating Numeric Role Fillers via Predictive Clustering Trees  for  Know...
Approximating Numeric Role Fillers via Predictive Clustering Trees for Know...
 
Inducing Predictive Clustering Trees for Datatype properties Values
Inducing Predictive Clustering Trees for Datatype properties ValuesInducing Predictive Clustering Trees for Datatype properties Values
Inducing Predictive Clustering Trees for Datatype properties Values
 
On the Effectiveness of Evidence-based Terminological Decision Trees
On the Effectiveness of Evidence-based Terminological Decision TreesOn the Effectiveness of Evidence-based Terminological Decision Trees
On the Effectiveness of Evidence-based Terminological Decision Trees
 
Inductive Classification through Evidence-based Models and Their Ensemble
Inductive Classification through Evidence-based Models and Their EnsembleInductive Classification through Evidence-based Models and Their Ensemble
Inductive Classification through Evidence-based Models and Their Ensemble
 
Tackling the Class Imbalance Learning Problem in Semantic Web Knowledge bases
Tackling the Class Imbalance Learning Problem in Semantic Web Knowledge basesTackling the Class Imbalance Learning Problem in Semantic Web Knowledge bases
Tackling the Class Imbalance Learning Problem in Semantic Web Knowledge bases
 
Towards Evidence Terminological Decision Tree
Towards Evidence Terminological Decision TreeTowards Evidence Terminological Decision Tree
Towards Evidence Terminological Decision Tree
 

Dernier

Pre-ProductionImproveddsfjgndflghtgg.pptx
Pre-ProductionImproveddsfjgndflghtgg.pptxPre-ProductionImproveddsfjgndflghtgg.pptx
Pre-ProductionImproveddsfjgndflghtgg.pptxStephen266013
 
How to Transform Clinical Trial Management with Advanced Data Analytics
How to Transform Clinical Trial Management with Advanced Data AnalyticsHow to Transform Clinical Trial Management with Advanced Data Analytics
How to Transform Clinical Trial Management with Advanced Data AnalyticsBrainSell Technologies
 
如何办理澳洲悉尼大学毕业证(USYD毕业证书)学位证成绩单原版一比一
如何办理澳洲悉尼大学毕业证(USYD毕业证书)学位证成绩单原版一比一如何办理澳洲悉尼大学毕业证(USYD毕业证书)学位证成绩单原版一比一
如何办理澳洲悉尼大学毕业证(USYD毕业证书)学位证成绩单原版一比一hwhqz6r1y
 
一比一原版纽卡斯尔大学毕业证成绩单如何办理
一比一原版纽卡斯尔大学毕业证成绩单如何办理一比一原版纽卡斯尔大学毕业证成绩单如何办理
一比一原版纽卡斯尔大学毕业证成绩单如何办理cyebo
 
How I opened a fake bank account and didn't go to prison
How I opened a fake bank account and didn't go to prisonHow I opened a fake bank account and didn't go to prison
How I opened a fake bank account and didn't go to prisonPayment Village
 
Data Visualization Exploring and Explaining with Data 1st Edition by Camm sol...
Data Visualization Exploring and Explaining with Data 1st Edition by Camm sol...Data Visualization Exploring and Explaining with Data 1st Edition by Camm sol...
Data Visualization Exploring and Explaining with Data 1st Edition by Camm sol...ssuserf63bd7
 
Toko Jual Viagra Asli Di Malang 081229400522 COD Obat Kuat Viagra Malang
Toko Jual Viagra Asli Di Malang 081229400522 COD Obat Kuat Viagra MalangToko Jual Viagra Asli Di Malang 081229400522 COD Obat Kuat Viagra Malang
Toko Jual Viagra Asli Di Malang 081229400522 COD Obat Kuat Viagra Malangadet6151
 
如何办理滑铁卢大学毕业证(Waterloo毕业证)成绩单本科学位证原版一比一
如何办理滑铁卢大学毕业证(Waterloo毕业证)成绩单本科学位证原版一比一如何办理滑铁卢大学毕业证(Waterloo毕业证)成绩单本科学位证原版一比一
如何办理滑铁卢大学毕业证(Waterloo毕业证)成绩单本科学位证原版一比一0uyfyq0q4
 
Generative AI for Trailblazers_ Unlock the Future of AI.pdf
Generative AI for Trailblazers_ Unlock the Future of AI.pdfGenerative AI for Trailblazers_ Unlock the Future of AI.pdf
Generative AI for Trailblazers_ Unlock the Future of AI.pdfEmmanuel Dauda
 
一比一原版加利福尼亚大学尔湾分校毕业证成绩单如何办理
一比一原版加利福尼亚大学尔湾分校毕业证成绩单如何办理一比一原版加利福尼亚大学尔湾分校毕业证成绩单如何办理
一比一原版加利福尼亚大学尔湾分校毕业证成绩单如何办理pyhepag
 
Data Analytics for Digital Marketing Lecture for Advanced Digital & Social Me...
Data Analytics for Digital Marketing Lecture for Advanced Digital & Social Me...Data Analytics for Digital Marketing Lecture for Advanced Digital & Social Me...
Data Analytics for Digital Marketing Lecture for Advanced Digital & Social Me...Valters Lauzums
 
如何办理新加坡国立大学毕业证(NUS毕业证)学位证成绩单原版一比一
如何办理新加坡国立大学毕业证(NUS毕业证)学位证成绩单原版一比一如何办理新加坡国立大学毕业证(NUS毕业证)学位证成绩单原版一比一
如何办理新加坡国立大学毕业证(NUS毕业证)学位证成绩单原版一比一hwhqz6r1y
 
edited gordis ebook sixth edition david d.pdf
edited gordis ebook sixth edition david d.pdfedited gordis ebook sixth edition david d.pdf
edited gordis ebook sixth edition david d.pdfgreat91
 
ℂall Girls Kashmiri Gate ℂall Now Chhaya ☎ 9899900591 WhatsApp Number 24/7
ℂall Girls Kashmiri Gate ℂall Now Chhaya ☎ 9899900591 WhatsApp  Number 24/7ℂall Girls Kashmiri Gate ℂall Now Chhaya ☎ 9899900591 WhatsApp  Number 24/7
ℂall Girls Kashmiri Gate ℂall Now Chhaya ☎ 9899900591 WhatsApp Number 24/7komalsharmaa480
 
社内勉強会資料  Mamba - A new era or ephemeral
社内勉強会資料   Mamba - A new era or ephemeral社内勉強会資料   Mamba - A new era or ephemeral
社内勉強会資料  Mamba - A new era or ephemeralNABLAS株式会社
 
Genuine love spell caster )! ,+27834335081) Ex lover back permanently in At...
Genuine love spell caster )! ,+27834335081)   Ex lover back permanently in At...Genuine love spell caster )! ,+27834335081)   Ex lover back permanently in At...
Genuine love spell caster )! ,+27834335081) Ex lover back permanently in At...BabaJohn3
 
1:1原版定制利物浦大学毕业证(Liverpool毕业证)成绩单学位证书留信学历认证
1:1原版定制利物浦大学毕业证(Liverpool毕业证)成绩单学位证书留信学历认证1:1原版定制利物浦大学毕业证(Liverpool毕业证)成绩单学位证书留信学历认证
1:1原版定制利物浦大学毕业证(Liverpool毕业证)成绩单学位证书留信学历认证ppy8zfkfm
 
一比一原版西悉尼大学毕业证成绩单如何办理
一比一原版西悉尼大学毕业证成绩单如何办理一比一原版西悉尼大学毕业证成绩单如何办理
一比一原版西悉尼大学毕业证成绩单如何办理pyhepag
 
Atlantic Grupa Case Study (Mintec Data AI)
Atlantic Grupa Case Study (Mintec Data AI)Atlantic Grupa Case Study (Mintec Data AI)
Atlantic Grupa Case Study (Mintec Data AI)Jon Hansen
 
Formulas dax para power bI de microsoft.pdf
Formulas dax para power bI de microsoft.pdfFormulas dax para power bI de microsoft.pdf
Formulas dax para power bI de microsoft.pdfRobertoOcampo24
 

Dernier (20)

Pre-ProductionImproveddsfjgndflghtgg.pptx
Pre-ProductionImproveddsfjgndflghtgg.pptxPre-ProductionImproveddsfjgndflghtgg.pptx
Pre-ProductionImproveddsfjgndflghtgg.pptx
 
How to Transform Clinical Trial Management with Advanced Data Analytics
How to Transform Clinical Trial Management with Advanced Data AnalyticsHow to Transform Clinical Trial Management with Advanced Data Analytics
How to Transform Clinical Trial Management with Advanced Data Analytics
 
如何办理澳洲悉尼大学毕业证(USYD毕业证书)学位证成绩单原版一比一
如何办理澳洲悉尼大学毕业证(USYD毕业证书)学位证成绩单原版一比一如何办理澳洲悉尼大学毕业证(USYD毕业证书)学位证成绩单原版一比一
如何办理澳洲悉尼大学毕业证(USYD毕业证书)学位证成绩单原版一比一
 
一比一原版纽卡斯尔大学毕业证成绩单如何办理
一比一原版纽卡斯尔大学毕业证成绩单如何办理一比一原版纽卡斯尔大学毕业证成绩单如何办理
一比一原版纽卡斯尔大学毕业证成绩单如何办理
 
How I opened a fake bank account and didn't go to prison
How I opened a fake bank account and didn't go to prisonHow I opened a fake bank account and didn't go to prison
How I opened a fake bank account and didn't go to prison
 
Data Visualization Exploring and Explaining with Data 1st Edition by Camm sol...
Data Visualization Exploring and Explaining with Data 1st Edition by Camm sol...Data Visualization Exploring and Explaining with Data 1st Edition by Camm sol...
Data Visualization Exploring and Explaining with Data 1st Edition by Camm sol...
 
Toko Jual Viagra Asli Di Malang 081229400522 COD Obat Kuat Viagra Malang
Toko Jual Viagra Asli Di Malang 081229400522 COD Obat Kuat Viagra MalangToko Jual Viagra Asli Di Malang 081229400522 COD Obat Kuat Viagra Malang
Toko Jual Viagra Asli Di Malang 081229400522 COD Obat Kuat Viagra Malang
 
如何办理滑铁卢大学毕业证(Waterloo毕业证)成绩单本科学位证原版一比一
如何办理滑铁卢大学毕业证(Waterloo毕业证)成绩单本科学位证原版一比一如何办理滑铁卢大学毕业证(Waterloo毕业证)成绩单本科学位证原版一比一
如何办理滑铁卢大学毕业证(Waterloo毕业证)成绩单本科学位证原版一比一
 
Generative AI for Trailblazers_ Unlock the Future of AI.pdf
Generative AI for Trailblazers_ Unlock the Future of AI.pdfGenerative AI for Trailblazers_ Unlock the Future of AI.pdf
Generative AI for Trailblazers_ Unlock the Future of AI.pdf
 
一比一原版加利福尼亚大学尔湾分校毕业证成绩单如何办理
一比一原版加利福尼亚大学尔湾分校毕业证成绩单如何办理一比一原版加利福尼亚大学尔湾分校毕业证成绩单如何办理
一比一原版加利福尼亚大学尔湾分校毕业证成绩单如何办理
 
Data Analytics for Digital Marketing Lecture for Advanced Digital & Social Me...
Data Analytics for Digital Marketing Lecture for Advanced Digital & Social Me...Data Analytics for Digital Marketing Lecture for Advanced Digital & Social Me...
Data Analytics for Digital Marketing Lecture for Advanced Digital & Social Me...
 
如何办理新加坡国立大学毕业证(NUS毕业证)学位证成绩单原版一比一
如何办理新加坡国立大学毕业证(NUS毕业证)学位证成绩单原版一比一如何办理新加坡国立大学毕业证(NUS毕业证)学位证成绩单原版一比一
如何办理新加坡国立大学毕业证(NUS毕业证)学位证成绩单原版一比一
 
edited gordis ebook sixth edition david d.pdf
edited gordis ebook sixth edition david d.pdfedited gordis ebook sixth edition david d.pdf
edited gordis ebook sixth edition david d.pdf
 
ℂall Girls Kashmiri Gate ℂall Now Chhaya ☎ 9899900591 WhatsApp Number 24/7
ℂall Girls Kashmiri Gate ℂall Now Chhaya ☎ 9899900591 WhatsApp  Number 24/7ℂall Girls Kashmiri Gate ℂall Now Chhaya ☎ 9899900591 WhatsApp  Number 24/7
ℂall Girls Kashmiri Gate ℂall Now Chhaya ☎ 9899900591 WhatsApp Number 24/7
 
社内勉強会資料  Mamba - A new era or ephemeral
社内勉強会資料   Mamba - A new era or ephemeral社内勉強会資料   Mamba - A new era or ephemeral
社内勉強会資料  Mamba - A new era or ephemeral
 
Genuine love spell caster )! ,+27834335081) Ex lover back permanently in At...
Genuine love spell caster )! ,+27834335081)   Ex lover back permanently in At...Genuine love spell caster )! ,+27834335081)   Ex lover back permanently in At...
Genuine love spell caster )! ,+27834335081) Ex lover back permanently in At...
 
1:1原版定制利物浦大学毕业证(Liverpool毕业证)成绩单学位证书留信学历认证
1:1原版定制利物浦大学毕业证(Liverpool毕业证)成绩单学位证书留信学历认证1:1原版定制利物浦大学毕业证(Liverpool毕业证)成绩单学位证书留信学历认证
1:1原版定制利物浦大学毕业证(Liverpool毕业证)成绩单学位证书留信学历认证
 
一比一原版西悉尼大学毕业证成绩单如何办理
一比一原版西悉尼大学毕业证成绩单如何办理一比一原版西悉尼大学毕业证成绩单如何办理
一比一原版西悉尼大学毕业证成绩单如何办理
 
Atlantic Grupa Case Study (Mintec Data AI)
Atlantic Grupa Case Study (Mintec Data AI)Atlantic Grupa Case Study (Mintec Data AI)
Atlantic Grupa Case Study (Mintec Data AI)
 
Formulas dax para power bI de microsoft.pdf
Formulas dax para power bI de microsoft.pdfFormulas dax para power bI de microsoft.pdf
Formulas dax para power bI de microsoft.pdf
 

Boosting dl concept learners

  • 1. Boosting DL Concept Learners Nicola FANIZZI, Giuseppe RIZZO and Claudia d'AMATO LACAM — Dipartimento di Informatica & CILA Università degli studi di Bari Aldo Moro ESWC 2019 — June 4th, 2019
  • 2. Outline Introduction The Framework: Training Strong learner Weak learner Test Empirical Evaluation Conclusions and Future Work
  • 4. Context & Motivations Considering knowledge bases in the Web of Data Goal: Inducing a classification function (predictive model) for a target concept to assess the membership of (unseen) individuals Supervised approach: each (training) example labeled as positive/negative/uncertain Possible Applications: approximate classification, semi-automatic KB completion (assertional part)
  • 5. Context & Motivations Several approaches: statistical approaches ⟶ prediction concept learners ⟶ prediction + intensional definition Poor model generalization (limited predictiveness)⚠
  • 6. Context & Motivations Weak Model 1 Strong Model Weak Model 2 ⋮ ⋮ ⋮ ⋮ Weak Model n Strong model more predictive than weak models Ensemble Classification (boosting, bagging,...)ⓘ
  • 7. Context & Motivations Contribution A framework for confidence-rated boosting for DL concept learners pool of weak models as concepts with the related score each model induced to cover hardest examples (leading to misclassifications)
  • 8. Example Learn a classification function for the individuals w.r.t. the target concept . Training set : (instances of with a known child), (instances of ) Ensemble Model: Father K = { { Man ⊑ Person, Woman ⊑ Person, Artist ⊑ Person, Man ⊑ ¬Woman} ∪ Man(a), Man(b), Man(c), Woman(d), Woman(f), Artist(e), Robot(g), hasChild(a, d), hasChild(b, e)} T Ps = {a, b} Man Ns = {d, f} Woman Us = {c, e, g} {Person [0.4], ¬Woman [0.8], Person [0.5], ∃hasChild. ⊤ [0.9]}
  • 10. Proposed Solution Given and a set of training examples Weight assigned to example (init. ) Ensemble of weak models (concepts) induced by a weak learner ( DLF): Confidence computed per weak model Weights updated according to the coverage of the hypothesis concept decreased for examples covered by increased for examples uncovered by K T wi ai = 1/|T|w (0) i T Ct ct Ct Ct
  • 11. Coverage: Definition An example covered by when i.e., is correctly classified x Ct ⎧ ⎩ ⎨ K ⊨ (x) if x ∈ PsCt K ⊨ ¬ (x) if x ∈ NsCt K ⊭ (x) ∧ K ⊭ ¬ (x) if x ∈ UsCt Ct x
  • 12. wDLF: Training Given , weights associated to the examples, and a partial description ( ) specialized via ref. operator ⟶ a set of concepts score: difference between sum of the weights of the covered ex.s and uncovered ex.s heuristic: maximum difference between the scores of and (gain) T w (t) Ct = ⊤Ct Ct ρ ⊑C ′ t Ct −∑ i| coveredai wi − −−−−−−−− √ ∑ i| uncoveredai wi − −−−−−−−−−− √ C ′ t Ct
  • 13. Weight Updates After is returned, weights of the examples updated: for each example the new weight will be where with depending on the (ratio between the)sum of the weights of covered and the uncovered (optimal value to minimize the training error) update Multiplied by for uncovered examples Divided by for covered examples Ct ai w (t) i = exp(− )w (t) i w (t) i ct = − ⋅ (−1ct α^t ) I( , ( ))li ht ei α^ exp ct exp ct
  • 14. Prediction Given the model and a query individual Prediction of single weak model : if if otherwise Final prediction: label according to weighted majority vote sums of confidence values for the three labels { [ ]}Ct ct T t=1 a Ct +1 K ⊨ (a)Ct −1 K ⊨ ¬ (a)Ct 0 =l ∗ argmax l∈{+1,−1,0} ∑ t∈T p( (x)=l)h t ct
  • 16. Evaluation: Settings 5 KBs for producing 15 artificial problems/target concepts (per KB) and related training & test sets Parameters: Ensemble size: refinements generated via wDLF: Baseline: wDLF Other systems: C and DL-F Boostrap procedure: 30 replications per learning problems/target Evaluation indices: match: exact predictions commission: opposite predictions omission: non definite predictions [for -label exs] induction: definite predictions [for -label exs] 10 30 ±1 0
  • 18. Evaluation: Examples of Induced Models
  • 20. Conclusions A boosted ensemble classifier how to derive a weak learner good performance for complex target concepts
  • 21. Future Work other (weak) learners exploiting ref.ops dynamic ensemble dimension (validation set) approximate reasoners for efficiency formatted by Markdeep 1.04 ✒