3/18/23, 6:49 PM Machine Learning,ML - - Unit 6 - Week 5 : Learning enabled by Prior Theories

(https://swayam.gov.in) (https://swayam.gov.in/nc_details/NPTEL)

psanthoshkumar.id@gmail.com 

NPTEL (https://swayam.gov.in/explorer?ncCode=NPTEL) » Machine Learning,ML (course)

Course
outline Week 5 : Assignment 5
The due date for submitting this assignment has passed.
How does an Due on 2023-03-01, 23:59 IST.
NPTEL online As per our records you have not submitted this assignment.
course work?
() 1) Explanation Based Learning typically takes a predefined Domain Theory and modifies 1 point
it in such a way that new more complex rules are created as specific combinations of several
Week 1: existing rules. The goal is to make problemsolving more efficient and the rule formation process is
Introduction guided by the problem examples considered for training. As part of the generalization process for
to the the newly constructed rules a specific kind of pattern matching technique is used. Which is the
Machine established term for this technique?
Learning
course () Variable matching
Resolution
Week 2: Recursion
Characterization
Unification
of Learning
Problems () No, the answer is incorrect.
Score: 0
Accepted Answers:
Week 3 :
Unification
Forms of
Representation
() 2) Domain Theory: 1 point
Safe to travel(P,A,C) :- Healthy(P), Error free(A), No epidemics (C).
Week 4 : Error free(A):- Passed an inspection (A). Errorfree(A):- Worked when driven same day (A).
Inductive Healthy(P) :- Had no symptoms for three weeks (P).
Learning Healthy(P) :- Passed a health test(P) Passed a health test (X):- On line test (X).
based on Passed a health test (X):- Clinic test (X). Passed an inspection (X):- Company inspection(X).
Symbolic Passed an inspection(X):- Authority inspection(X). No epidemics (C):-No WHO alert (C).
Representations Operational predicates:

https://onlinecourses.nptel.ac.in/noc23_cs11/unit?unit=51&assessment=101 1/6
3/18/23, 6:49 PM Machine Learning,ML - - Unit 6 - Week 5 : Learning enabled by Prior Theories

and Weak Worked when driven last same day, Had no symptoms for three weeks, On line test, Clinic test,
Theories () Company inspection, Authority inspection, No WHO alert
Training Example:
Week 5 : Safe to travel (John, IndiaAir2020, India) Passed a health test(John), Authority
Learning inspection(IndiaAir 2020), No WHO alert (India).
enabled by Goal concept:
Prior Safe to travel(P, A, C)
Theories ()
Explain and generalize from Safe To Travel( John, IndiaAir2020, India). How does the resulting rule
Machine
look like after applying an EBL algorithm
Learning
enabled by
Safe to travel(John,IndiaAir2020,India)) :- Healthy(John), Error free(IndiaAir2020), No
Prior Theories
(unit?
epidemics (India).
unit=51&lesson=52) Safe to travel(John, IndiaAir2020, India):-Passed a health test (John), Authority
inspection(IndiaAir 2020), No WHO alert (India).
Explanation
Based Learning Safe to travel(P, A, C):-Passed a health test (P), Authority inspection(A), No WHO alert (C).
(unit? Safe to travel(P,A,C)) :- Healthy(P), Error free(A), No epidemics (C).
unit=51&lesson=53)
No, the answer is incorrect.
Inductive Logic Score: 0
Programming Accepted Answers:
(unit? Safe to travel(P, A, C):-Passed a health test (P), Authority inspection(A), No WHO alert (C).
unit=51&lesson=54)
3) Which of the following inference schemes correspond to Induction? 1 point
Reinforcement
Learning - Part
01 Introduction
(unit?
unit=51&lesson=55)

Reinforcement
Learning - Part
02 Learning
Algorithms
(unit?
unit=51&lesson=56)

Reinforcement
Learning - Part
03 Q - Learning
(unit?
unit=51&lesson=57) A
B
Case - Based
Reasoning C
(unit? No, the answer is incorrect.
unit=51&lesson=58) Score: 0
Accepted Answers:
Lecture Notes
B
(unit?
unit=51&lesson=59)
4) ILP algorithm constructs a lattice of clauses fabricated from the building blocks 1 point
(predicates, variables, constants) specific to each induction case and navigates that lattice in order

https://onlinecourses.nptel.ac.in/noc23_cs11/unit?unit=51&assessment=101 2/6
3/18/23, 6:49 PM Machine Learning,ML - - Unit 6 - Week 5 : Learning enabled by Prior Theories

Quiz: Week 5 : to establish a Hypothesis set that fits the current set of examples. ILP can work in two fashions: Top
Assignment 5 Down and Bottom Up, establishing specialization or generalization relations in the lattice.
(assessment?
name=101) Case
Hypotheses: insect(X).
Week 5
Feedback form
Positive Examples:
: Machine insect(silverfish), insect(green drake), insect(cricket)
Learning, ML Background knowledge:
(unit? Has six legs (fire brat), Has six legs (green drake), Have wings(green drake), Has six legs
unit=51&lesson=60) (cricket), Have wings(cricket), Can fold wings(cricket),

Week 6 :
How does the Hypotheses set of clauses look like after a first iteration of a Top Down ILP algorithm
Machine
based on the first positive example insect(fire brat) ?
Learning
based insect(X):-Has six legs(X).
Artificial
insect(X):-Has six legs(X), insect(X):-Has wings(X), insect (X):-Can fold wings(X).
Neural
Networks () insect(fire brat):-Has six legs(fire brat).

No, the answer is incorrect.

Week 7 : Score: 0
Tools and Accepted Answers:
Resources insect(X):-Has six legs(X), insect(X):-Has wings(X), insect (X):-Can fold wings(X).
and
5) Instance Based Learning techniques are typically used in one of the main phases of 1 point
Interdisciplinary
Case Based Reasoning. Which?
Inspiration ()
RETRIEVE
Week 8 ()
REUSE

Lecture Notes RETAIN

() REVISE

No, the answer is incorrect.

Text Score: 0
Transcripts () Accepted Answers:
RETRIEVE
Download
Videos () 6) 1 point

Books ()

Lecture notes
()

Problem
Solving
Session ()

https://onlinecourses.nptel.ac.in/noc23_cs11/unit?unit=51&assessment=101 3/6
3/18/23, 6:49 PM Machine Learning,ML - - Unit 6 - Week 5 : Learning enabled by Prior Theories

Choose the correct sequence of phases:

X Y Z W = RETRIEVE REUSE RETAIN REVISE

X Y Z W = RETRIEVE REVISE REUSE RETAIN
X Y Z W = RETAIN REVISE REUSE RETRIEVE
X Y Z W = RETAIN RETRIEVE REVISE REUSE
No, the answer is incorrect.
Score: 0
Accepted Answers:
X Y Z W = RETAIN RETRIEVE REVISE REUSE

7) A dynamic programming approach to calculating the shortest distance from S to T gives1 point
the result= 14.

Which result would a greedy forward search give?

26
21
14

No, the answer is incorrect.

Score: 0
Accepted Answers:
26

8) Monte Carlo Reinforcement Learning MC methods learn directly from samples of 1 point
complete episodes of experience.
First-visit MC: average returns only for first time s is visited in an episode.
Every-Visit MC: average returns for every time s is visited in an episode.

Simplified Algorithm for first-visit Monte Carlo

1. Initialize state-value functions. Return list(s) ← empty list
2. For all s in all episodes E

https://onlinecourses.nptel.ac.in/noc23_cs11/unit?unit=51&assessment=101 4/6
3/18/23, 6:49 PM Machine Learning,ML - - Unit 6 - Week 5 : Learning enabled by Prior Theories

3. return = sum of rewards r in episode from state to termination.

4. if this is the first occurrence of this state s add the calculated return to the returns list (s).
5. Calculate values of all s as average over all return lists (s).

Example: An undiscounted Markov Reward Process with two states A and B. The Transition matrix
and reward function are unknown. Two sample episodes: E1 A->A, r=2 A->B, r=7 B->A, r= -7 A->B,
r=7 B->terminate, r= -2. E2 B->A, r= -7 A->B, r=7 B->terminate, r = -2

Which are the estimated values of A and B after considering the two episodes using the simplified
first-visit algorithm?

V(A)= 6,V(B)= -2
V(A)= 7,V(B)= -2
V(A)= 6,V(B)= 2
V(A)= 7,V(B)= 2
No, the answer is incorrect.
Score: 0
Accepted Answers:
V(A)= 6,V(B)= -2

9) 1 point

https://onlinecourses.nptel.ac.in/noc23_cs11/unit?unit=51&assessment=101 5/6
3/18/23, 6:49 PM Machine Learning,ML - - Unit 6 - Week 5 : Learning enabled by Prior Theories

In this example, we have a 4X5 board. The start state is marked above with a blue dot. The rewards
for all positions are also marked (rewards being 1, -4 and 4). The positions on the rim of the board
are terminal states. The board elements are labelled from 1 to 20 as in the mid figure. Moves are N,
S, E and W. We consider two episodes: 1. 12:E->13:E->14:E->15 and 2. 12:N->7:E->8:E->9:E->10.
What are the updated versions of the Q elements for state 12, 13, 14, 7,8 and 9 after considering
these two episodes? α = γ = 1 .

12,E=0 13,E=0, 14,E= -4 12,N=0, 7,E=0 8,E=0 9,E = 4

12,E=1 13,E=1 14,E= -6 12,N=1 7,E=1 8,E=1 9,E= 6
12,E=1 13,E=2 14,E= -6 12,N=1 7,E=2 8,E=3 9,E= 6
12,E=1 13,E=2 14,E= -2 12,N=1, 7,E=2 8,E=3 9,E=7

No, the answer is incorrect.

Score: 0
Accepted Answers:
12,E=1 13,E=1 14,E= -6 12,N=1 7,E=1 8,E=1 9,E= 6

https://onlinecourses.nptel.ac.in/noc23_cs11/unit?unit=51&assessment=101 6/6