using qualitative knowledge in numerical learning

84
USING QUALITATIVE KNOWLEDGE IN NUMERICAL LEARNING Ivan Bratko Faculty of Computer and Info. Sc. University of Ljubljana Slovenia

Upload: butest

Post on 08-May-2015

160 views

Category:

Documents


1 download

TRANSCRIPT

Page 1: Using Qualitative Knowledge in Numerical Learning

USING

QUALITATIVE KNOWLEDGE

IN NUMERICAL LEARNING

Ivan Bratko

Faculty of Computer and Info. Sc.

University of Ljubljana

Slovenia

Page 2: Using Qualitative Knowledge in Numerical Learning

THIS TALK IS ABOUT:

AUTOMATED MODELLING FROM DATA

WITH MACHINE LEARNING

COMBINING NUMERICAL AND QUALITATIVE

REPRESENTATIONS

Page 3: Using Qualitative Knowledge in Numerical Learning

3

BUILING MODELS FROM DATA

Observed

system

Machine learning,

numerical regression

Model of system

Data

Page 4: Using Qualitative Knowledge in Numerical Learning

EXAMPLE: POPULATION DYNAMICS

A lake with zooplankton, phytoplankton and nutrient nitrogen

Variables in system:

Nut

Phyto

Zoo

Page 5: Using Qualitative Knowledge in Numerical Learning

POPULATION DYNAMICS

Observed behaviour in time

0 20 40 60 800.0

0.4

0.8

1.2

1.6

NutPhytoZoo

Data provided by Todorovski&Džeroski

Page 6: Using Qualitative Knowledge in Numerical Learning

PRIOR KNOWLEDGE

We would like our modelling methods to make use of expert’s prior knowledge (possibly qualitative)

Phytoplankton feeds on Nutrient,

Zooplankton feeds on Phytoplankton

Nutrient Phyto Zoo

Page 7: Using Qualitative Knowledge in Numerical Learning

QUALITATIVE DIFFICULTIES OF

NUMERICAL LEARNING

Learn time behavior of water level:

h = f( t, initial_outflow)

Level h

outflow t

h

Page 8: Using Qualitative Knowledge in Numerical Learning

0

20

40

60

80

100

h

t=1

t=19

TIME BEHAVIOUR OF WATER LEVEL

Initial_ouflow=12.5

Page 9: Using Qualitative Knowledge in Numerical Learning

0

20

40

60

80

100

h

t=1

t=19

VARYING INITIAL OUTFLOW

Initial_ouflow=12.5

11.25

10.08.75

6.25

Page 10: Using Qualitative Knowledge in Numerical Learning

0

20

40

60

80

100

h

t=1

t=19

PREDICTING WATER LEVEL WITH M5

Qualitatively incorrect – water level cannot increase

M5 prediction

11.25

10.08.75

6.25

7.5Initial_ouflow

=12.5

Page 11: Using Qualitative Knowledge in Numerical Learning

QUALITATIVE ERRORS OF

NUMERICAL LEARNERS

Experiments with regression (model) trees (M5; Quinlan 92), LWR (Atkenson et.al. 97) in Weka (Witten & Frank 2000), neural nets, ...

Qualitative errors:

water level should never increase

water level should not be negative

An expert might accept numerical errors, but such qualitative errors are particularly disturbing

Page 12: Using Qualitative Knowledge in Numerical Learning

Q2 LEARNING

AIMS AT OVERCOMING THESE

DIFFICULTIES

Page 13: Using Qualitative Knowledge in Numerical Learning

Q2 LEARNINGŠuc, Vladušič, Bratko; IJCAI’03, AIJ 2004, IJCAI’05

Aims at overcoming these difficulties of numerical learning

Q2 = Qualitatively faithful Quantitative learning

Q2 makes use of qualitative constraints

Page 14: Using Qualitative Knowledge in Numerical Learning

QUALITATIVE CONSTRAINTS

FOR WATER LEVEL

For any initial outflow:

Level is always decreasing with time

For any time point:

Greater the initial outflow, greater the level

Page 15: Using Qualitative Knowledge in Numerical Learning

SUMMARY OF Q2 LEARNING

Standard numerical learning approaches make qualitative errors.

As a result, numerical predictions are qualitatively inconsistent with expectations

Q2 learning (Qualitatively faithful Quantitative prediction);

A method that enforces qualitative consistency

Resulting numerical models enable clearer interpretation, and also significantly improve quantitative prediction

Page 16: Using Qualitative Knowledge in Numerical Learning

IDEA OF Q2

First find qualitative laws in data

Respect these qualitative laws in numerical learning

Page 17: Using Qualitative Knowledge in Numerical Learning

CONTENTS OF REST OF TALK

Building blocks of Q2 learning:

Ideas from Qualitative Reasoning,

Algorithms QUIN, QFILTER, QCGRID

Experimental analysis

Applications:

Car modelling, ecological modelling, behavioural cloning (operating a crane, flying an aircraft)

Page 18: Using Qualitative Knowledge in Numerical Learning

HOW CAN WE DESCRIBE QUALITATIVE

PROPERTIES?

We can use concepts from field of qualitative

reasoning in AI

Related terms:

Qualitative physics, Naive physics, Qualitative

modelling

Page 19: Using Qualitative Knowledge in Numerical Learning

ESSENCE OF NAIVE PHYSICS

Describe physical processes qualitatively, without numbers or exact numerical relations

“Naive physics”, as opposed to "proper physics“

Close to common sense descriptions

Page 20: Using Qualitative Knowledge in Numerical Learning

EXAMPLE: BATH TUB

What will happen?

Amount of water will keep increasing,so will level,until the level reaches the top.

Page 21: Using Qualitative Knowledge in Numerical Learning

EXAMPLE: U-TUBE

What will happen?

La

Lb

Level La will be decreasing,and Lb increasing,until La = Lb.

Page 22: Using Qualitative Knowledge in Numerical Learning

QUALITATIVE REASONING ABOUT U-TUBE

Total amount of water in system constant

If La > Lb then flow from A to B

Flow causes amount in A to decrease

Flow causes amount in B to increase

All changes in time happen continuously and smoothly

Level La

Level Lb

A B

Page 23: Using Qualitative Knowledge in Numerical Learning

QUALITATIVE REASONING ABOUT U-TUBE

In any container: the greater the amount, the greater the

level

So, La will keep decreasing, Lb increasing

Level La

Level Lb

Page 24: Using Qualitative Knowledge in Numerical Learning

QUALITATIVE REASONING ABOUT U-TUBE

La will keep decreasing, Lb increasing, until they equalise

Level La

Level Lb

La

Lb

Time

Page 25: Using Qualitative Knowledge in Numerical Learning

THIS REASONING IS VALID FOR ALL

CONTAINERS

OF ANY SHAPE AND SIZE,

REGARDLESS OF ACTUAL NUMBERS!

Page 26: Using Qualitative Knowledge in Numerical Learning

QHY REASON QUALITATIVELY?

Because it is easier than quantitatively

Because it is easy to understand -

facilitates explanation

We want to exploit these advantages in ML

Page 27: Using Qualitative Knowledge in Numerical Learning

RELATION BETWEEN

AMOUNT AND LEVEL

The greater the amount, the greater the level

A = M+(L)

A is a monotonically increasing function of L

Page 28: Using Qualitative Knowledge in Numerical Learning

MONOTONIC FUNCTIONS

Y = M+(X) specifies a family of functions

X

Y

Page 29: Using Qualitative Knowledge in Numerical Learning

MONOTONIC QUALITATIVE CONSTRAINTS,

MQCs

Generalisation of monotonically increasing functions to several arguments

Example: Z = M+,- ( X, Y)

Z increases with X, and decreases with Y

More precisely: if X increases and Y stays unchanged then Z increases

Page 30: Using Qualitative Knowledge in Numerical Learning

EXAMPLE: BEHAVIOUR OF GAS

Pressure = M+,- (Temperature, Volume)

Pressure increases with Temperature

Pressure decreaseswith Volume

Page 31: Using Qualitative Knowledge in Numerical Learning

Q2 LEARNING

Induce qualitative constraints(QUIN)

Qualitative to Quantitative Transformation (Q2Q)

Numerical predictor: • respects qualitative constraints • fits data numerically

Numerical dataOne possibility: QFILTER

Page 32: Using Qualitative Knowledge in Numerical Learning

PROGRAM QUIN

INDUCING QUALITATIVE CONSTRAINTS

FROM NUMERICAL DATA

Šuc 2001 (PhD Thesis, also as book 2003)

Šuc and Bratko, ECML’01

Page 33: Using Qualitative Knowledge in Numerical Learning

QUIN

QUIN = Qualitative Induction

Numerical examples

QUIN

Qualitative tree

Qualitative tree: similar to decision tree,

qualitative constraints in leaves

Page 34: Using Qualitative Knowledge in Numerical Learning

EXAMPLE PROBLEM FOR QUIN

Noisy examples:

z = x 2 - y 2 + noise(st.dev. 50)

Page 35: Using Qualitative Knowledge in Numerical Learning

EXAMPLE PROBLEM FOR QUIN

In this region:z = M+,+(x,y)

Page 36: Using Qualitative Knowledge in Numerical Learning

INDUCED QUALITATIVE TREE FOR

z = x2 - y2 + noise

z=M-,+

(x,y) z=M-,-

(x,y) z=M+,+

(x,y) z=M+,-

(x,y)

0> 0 > 0

0 > 0

0

y

x

y

Page 37: Using Qualitative Knowledge in Numerical Learning

QUIN ALGORITHM: OUTLINE

Top-down greedy algorithm (similar to induction of decision trees)

For every possible split, find the “most consistent” MQC (min. error-cost) for each subset of examples

Select the best split according to MDL

Page 38: Using Qualitative Knowledge in Numerical Learning

Q2Q

Qualitative to Quantitative Transformation

Page 39: Using Qualitative Knowledge in Numerical Learning

Q2Q EXAMPLE

X < 5

y n

Y = M+(X) Y = M-(X)

5 X

Y

Page 40: Using Qualitative Knowledge in Numerical Learning

QUALITATIVE TREES IMPOSE

NUMERICAL CONSTRAINTS

MQCs impose numerical constraints on class

values, between pairs of examples

y = M+(x) requires:

If x1 > x2 then y1 > y2

Page 41: Using Qualitative Knowledge in Numerical Learning

RESPECTING MQCs NUMERICALLY

z = M+,+(x,y) requires:

If x1 < x2 and y1 < y2 then z1 < z2

(x2, y2)

(x1, y1)

x

y

Page 42: Using Qualitative Knowledge in Numerical Learning

QFILTER

AN APPROACH TO Q2Q

TRANSFORMATION

Šuc and Bratko, ECML’03

Page 43: Using Qualitative Knowledge in Numerical Learning

TASK OF QFILTER

Given: qualitative tree points with class predictions by arbitrary numerical

learner learning examples (optionally)

Modify class predictions to achieve consistency with qualitative tree

Page 44: Using Qualitative Knowledge in Numerical Learning

QFILTER IDEA

Force numerical predictions to respect

qualitative constraints:

find minimal changes of predicted values so that qualitative constraints become satisfied

“minimal” = min. sum of squared changes

a quadratic programming problem

Page 45: Using Qualitative Knowledge in Numerical Learning

RESPECTING MQCs NUMERICALLY

Y = M+(X)

X

Y

Page 46: Using Qualitative Knowledge in Numerical Learning

QFILTER APPLIED TO WATER OUTFLOW

Qualitative constraint that applies to

water outflow:

h = M -,+(time, InitialOutflow)

This could be supplied by domain expert,

or induced from data by QUIN

Page 47: Using Qualitative Knowledge in Numerical Learning

0

20

40

60

80

100

h

t=1

t=19

PREDICTING WATER LEVEL WITH M5

M5 prediction

7.5

Page 48: Using Qualitative Knowledge in Numerical Learning

QFILTER’S PREDICTION

QFILTER predictions

True values

Page 49: Using Qualitative Knowledge in Numerical Learning

POPULATION DYNAMICS

Aquatic ecosystem with zooplankton, phytoplankton and nutrient nitrogen

Phyto feeds on Nutrient,

Zoo feeds on Phyto

Nutrient Phyto Zoo

Page 50: Using Qualitative Knowledge in Numerical Learning

POPULATION DYNAMICS WITH Q2

Behaviour in time

0 20 40 60 800.0

0.4

0.8

1.2

1.6

NutPhytoZoo

Page 51: Using Qualitative Knowledge in Numerical Learning

PREDICTION PROBLEM

Predict the change in zooplankton population:

ZooChange(t) = Zoo(t + 1) - Zoo(t)

Biologist’s rough idea:

ZooChange = Growth - Mortality

M+,+(Zoo, Phyto) M+(Zoo)

Page 52: Using Qualitative Knowledge in Numerical Learning

APPROXIMATE QUALITATIVE MODEL

OF ZOO CHANGE

Induced from data by QUIN

Page 53: Using Qualitative Knowledge in Numerical Learning

EXPERIMENT WITH NOISY DATA

Domain no noise

LWR; Q2

5 % noise

LWR; Q2

20 % noise

LWR; Q2

ZooChange 0.015 ; 0.008 0.112 ; 0.102 2.269 ; 1.889

All results as MSE (Mean Squared Error)

Page 54: Using Qualitative Knowledge in Numerical Learning

APPLICATIONS OF Q2

FROM REAL ECOLOGICAL DATA

Growth of algae Lagoon of Venice

Plankton in Lake Glumsoe

Page 55: Using Qualitative Knowledge in Numerical Learning

Lake Glumsø

Location and properties:

Lake Glumsø is located in a sub-glacial valley in Denmark

Average depth 2 m

Surface area 266000 m2

Pollution

Receives waste water from community with 3000 inhabitants (mainly agricultural)

High nitrogen and phosphorus concentration in waste water caused hypereutrophication

No submerged vegetation

low transparency of water

oxygen deficit at the bottom of the lake

Page 56: Using Qualitative Knowledge in Numerical Learning

Lake Glumsø – data

Relevant variables for modelling are:

phytoplankton phyto

zooplankton zoo

soluble nitrogen ns

soluble phosphorus ps

water temperature temp

Page 57: Using Qualitative Knowledge in Numerical Learning

PREDICTION ACCURACY

• Over all (40) experiments.• Q2 better than LWR in 75% (M5, 83%) of the test cases• The differences were found significant (t-test)

at 0.02 significance level

Page 58: Using Qualitative Knowledge in Numerical Learning

OTHER ECOLOGICAL MODELLING

APPLICATIONS

Predicting ozone concentrations in Ljubljana and Nova Gorica

Predicting flooding of Savinja river

Q2 model by far superior to any predictor so far used in practice

Page 59: Using Qualitative Knowledge in Numerical Learning

CASE STUDY

INTEC’S CAR SIMULATION MODELS

Goal: simplify INTEC’s car models to speed up simulation

Context: Clockwork European project (engineering design)

Page 60: Using Qualitative Knowledge in Numerical Learning

Intec’s wheel model

Page 61: Using Qualitative Knowledge in Numerical Learning

WHEEL MODEL: PREDICTING TOE ANGLE

10 30 50 70 90

-0.010

-0.006

-0.002

0.002

alpha

time in steps dt=0.7 sec.

Page 62: Using Qualitative Knowledge in Numerical Learning

WHEEL MODEL: PREDICTING TOE ANGLE

10 30 50 70 90

-0.010

-0.006

-0.002

0.002

alphaLWR predicted alpha

time in steps dt=0.7 sec.

Page 63: Using Qualitative Knowledge in Numerical Learning

WHEEL MODEL: PREDICTING TOE ANGLE

10 30 50 70 90

-0.010

-0.006

-0.002

0.002

alphaLWR predicted alphaM5 predicted alpha

time in steps dt=0.7 sec.

Page 64: Using Qualitative Knowledge in Numerical Learning

10 30 50 70 90

-0.010

-0.006

-0.002

0.002

alphaLWR predicted alphaM5 predicted alphaM5 predicted alpha

time in steps dt=0.7 sec.

WHEEL MODEL: PREDICTING TOE ANGLE

Q2

Qualiative errors

Q2 predicted alpha

Page 65: Using Qualitative Knowledge in Numerical Learning

BEHAVIOURAL CLONING

Given a skilled operator, reconstruct the human’s sub cognitive skill

Page 66: Using Qualitative Knowledge in Numerical Learning

EXAMPLE: GANTRY CRANE

Control force

Load

Carriage

Page 67: Using Qualitative Knowledge in Numerical Learning

USE MACHINE LEARNING:

BASIC IDEA

Controller

System

Observe

Execution trace

Learning program

Reconstructed controller (“clone”)

ActionsStates

Page 68: Using Qualitative Knowledge in Numerical Learning

CRITERIA OF SUCCESS

Induced controller description has to:

Be comprehensible

Work as a controller

Page 69: Using Qualitative Knowledge in Numerical Learning

WHY COMPREHENSIBILITY?

To help the user’s intuition about the

essential mechanism and

causalities

that enable the controller achieve the goal

Page 70: Using Qualitative Knowledge in Numerical Learning

SKILL RECONSTRUTION IN CRANE

X0=0L0=20

load

trolley

X

L

Xg=60Lg=32

Control forces: Fx, FL

State: X, dX, , d, L, dL

Page 71: Using Qualitative Knowledge in Numerical Learning

CARRIAGE CONTROL

QUIN: dXdes= f(X, , d)

M-(X) M+()

X < 20.7

X < 60.1M+(X)

yes

yes

no

no

First the trolley velocity is increasing

First the trolley velocity is increasing

From about middle distance from the goal until the goal the trolley velocity is decreasing

From about middle distance from the goal until the goal the trolley velocity is decreasing

At the goal reduce the swing of the rope (by acceleration of the trolley when the rope angle increases)

At the goal reduce the swing of the rope (by acceleration of the trolley when the rope angle increases)

Page 72: Using Qualitative Knowledge in Numerical Learning

CARRIAGE CONTROL: dXdes= f(X, , d)

M-(X) M+()

X < 20.7

X < 60.1

X < 29.3

M+(X) d < -0.02

M-(X) M-,+(X,)

M+,+,-(X, , d)

yes

yes

yes

yes

no

no

no

no

Enables reconstruction of

individual differences in control styles

Operator S Operator L

Page 73: Using Qualitative Knowledge in Numerical Learning

CASE STUDY IN REVERSE

ENGINEERING:

ANTI-SWAY CRANE

Page 74: Using Qualitative Knowledge in Numerical Learning

ANTI-SWAY CRANE

Industrial crane controller minimising load swing, “anti-sway crane”

Developed by M. Valasek (Czech Technical University, CTU)

Reverse engineering of anti-sway crane: a case study in the Clockwork European project

Page 75: Using Qualitative Knowledge in Numerical Learning

ANTI-SWAY CRANE OF CTU

Crane parameters:

travel distance 100m

height 15m, width 30m

80-120 tons

In daily use at Nova Hut metallurgical factory, Ostrava

Page 76: Using Qualitative Knowledge in Numerical Learning

EXPLAINING

HOW CONTROLLER WORKS

Load swinging to right;

Accelerate cart to right to reduce swing

Page 77: Using Qualitative Knowledge in Numerical Learning

EMPIRICAL EVALUATION

Compare errors of base-learners and corresponding Q2 learners

differences btw. a base-learner and a Q2 learner are only due to the induced qualitative constraints

Experiments with three base-learners:

Locally Weighted Regression (LWR)

Model trees

Regression trees

Page 78: Using Qualitative Knowledge in Numerical Learning

Y1 Y2

Robot Arm Domain

Two-link, two-joint robot arm

Link 1 extendible: L1 [2, 10]

Y1 = L1 sin(1)

Y2 = L1 sin(1) + 5 sin(1+2) 1

2

Four learning problems:

A: Y1 = f(L1, 1)

B: Y2 = f(L1, 1, 2, sum, Y1)

C: Y2 = f(L1, 1, 2 , sum)

D: Y2 = f(L1, 1, 2)

L1

Derived attribute

sum= 1 + 2

Difficulty for Q2

Page 79: Using Qualitative Knowledge in Numerical Learning

Robot Arm: LWR and Q2 at different noise levels

0.1

0.3

0.5

A 0%n. A 5%n. A 10%n B 0%n. B 5%n. B 10%n C 0%n. C 5%n. C 10%n D 0%n. D 5%n. D 10%n

RR

E

LWR

Q2+LWR

Q2 outperforms LWR with all four learning problems (at all three noise levels)

A 0, 5, 10% n.| B 0, 5, 10% n.| C 0, 5, 10% n.| D 0, 5, 10% n.

Page 80: Using Qualitative Knowledge in Numerical Learning

UCI and Dynamic Domains

Five smallest regression data sets from UCI

Dynamic domains: typical domains where QUIN was applied so far to explain

the control skill or control the system until now was not possible to measure accuracy of the

learned concepts (qualitative trees)

AntiSway logged data from an anti-sway crane controller

CraneSkill1, CraneSkill2: logged data of experienced human operators controlling a

crane

Page 81: Using Qualitative Knowledge in Numerical Learning

UCI and Dynamic Domains: LWR compared to Q2

0

0.1

0.2

0.3

0.4

0.5

0.6

AutoMpg AutoPrice Housing Mach.CPU Servo CraneSkil1 CraneSkill2 AntiSway

RR

E

LWR Q2+LWR

Similar results with other two base-learners. Q2

significantly better than base-learners in 18 out of 24 comparisons (24 = 8 datasets * 3 base-learners)

Page 82: Using Qualitative Knowledge in Numerical Learning

Q2 - CONCLUSIONS

A novel approach to numerical learning

Can take into account qualitative prior knowledge

Advantages:

qualitative consistency of induced models and data – important for interpretation of induced models

improved numerical accuracy of predictions

Page 83: Using Qualitative Knowledge in Numerical Learning

Q2 TEAM + ACKNOWLEDGEMENTS

Q2 learning, QUIN, Qfilter, QCGRID (AI Lab, Ljubljana):

Dorian Šuc

Daniel Vladušič

Car modelling data

Wolfgan Rulka (INTEC, Munich)

Zbinek Šika (Czech Technical Univ.)

Population dynamics data

Sašo Džeroski, Ljupčo Todorovski (J. Stefan Institute, Ljubljana)

Lake Glumsoe

Sven Joergensen

Boris Kompare, Jure Žabkar, D. Vladušič

Page 84: Using Qualitative Knowledge in Numerical Learning

RELEVANT PAPERS

Clark and Matwin 93: also used qualitative constraints in numerical predictions

Šuc, Vladušič and Bratko; IJCAI’03

Šuc, Vladušič and Bratko; Artificial Intelligence Journal, 2004

Šuc and Bratko; ECML’03

Šuc and Bratko; IJCAI’05