4.

G REEDY A LGORITHMS I

‣ coin changin
‣ interval schedulin
‣ interval partitionin
‣ scheduling to minimize lateness
g

 4. G REEDY A LGORITHMS I

‣ coin changin
‣ interval schedulin
‣ interval partitionin
‣ scheduling to minimize lateness
g

Coin changing

Goal. Given U. S. currency denominations { 1, 5, 10, 25, 100 },

devise a method to pay amount to customer using fewest coins.

Ex. 34¢.

3
Coin changing

Goal. Given U. S. currency denominations { 1, 5, 10, 25, 100 },

devise a method to pay amount to customer using fewest coins.

Ex. 34¢.

3
Coin changing

Goal. Given U. S. currency denominations { 1, 5, 10, 25, 100 },

devise a method to pay amount to customer using fewest coins.

Ex. 34¢.

Cashier′s algorithm. At each iteration, add coin of the largest value that does not
take us past the amount to be paid.

3
Coin changing

Goal. Given U. S. currency denominations { 1, 5, 10, 25, 100 },

devise a method to pay amount to customer using fewest coins.

Ex. 34¢.

Cashier′s algorithm. At each iteration, add coin of the largest value that does not
take us past the amount to be paid.

Ex. $2.89.

3
Coin changing

Goal. Given U. S. currency denominations { 1, 5, 10, 25, 100 },

devise a method to pay amount to customer using fewest coins.

Ex. 34¢.

Cashier′s algorithm. At each iteration, add coin of the largest value that does not
take us past the amount to be paid.

Ex. $2.89.

3
Cashier′s algorithm

At each iteration, add coin of the largest value that does not take us past the amount
to be paid.

CASHIERS-ALGORITHM (x, c1, c2, …, cn)

_________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________

SORT n coin denominations so that 0 < c1 < c2 < … < cn.

S ← ∅. multiset of coins selected

WHILE (x > 0)
k ← largest coin denomination ck such that ck ≤ x.
IF (no such k)
RETURN “no solution.
ELSE
x ← x – ck.
S ← S ∪ { k }.
RETURN S.
________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________

4
”

Greedy algorithms I: quiz 1

Is the cashier’s algorithm optimal

A. Yes, greedy algorithms are always optimal.

B. Yes, for any set of coin denominations c1 < c2 < … < cn provided c1 = 1.

C. Yes, because of special properties of U.S. coin denominations

D. No.

5
?

Greedy algorithms I: quiz 1

Is the cashier’s algorithm optimal

A. Yes, greedy algorithms are always optimal.

B. Yes, for any set of coin denominations c1 < c2 < … < cn provided c1 = 1.

C. Yes, because of special properties of U.S. coin denominations

D. No.

5
?

Cashier′s algorithm (for arbitrary coin denominations)

Q. Is cashier’s algorithm optimal for any set of denominations?

6
Cashier′s algorithm (for arbitrary coin denominations)

Q. Is cashier’s algorithm optimal for any set of denominations?

A. No. Consider U.S. postage: 1, 10, 21, 34, 70, 100, 350, 1225, 1500.

6
Cashier′s algorithm (for arbitrary coin denominations)

Q. Is cashier’s algorithm optimal for any set of denominations?

A. No. Consider U.S. postage: 1, 10, 21, 34, 70, 100, 350, 1225, 1500.
・Cashier’s algorithm: 140¢ = 100 + 34 + 1 + 1 + 1 + 1 + 1 + 1.

6
Cashier′s algorithm (for arbitrary coin denominations)

Q. Is cashier’s algorithm optimal for any set of denominations?

A. No. Consider U.S. postage: 1, 10, 21, 34, 70, 100, 350, 1225, 1500.
・Cashier’s algorithm: 140¢ = 100 + 34 + 1 + 1 + 1 + 1 + 1 + 1.
・Optimal: 140¢ = 70 + 70.

6
Cashier′s algorithm (for arbitrary coin denominations)

Q. Is cashier’s algorithm optimal for any set of denominations?

A. No. Consider U.S. postage: 1, 10, 21, 34, 70, 100, 350, 1225, 1500.
・Cashier’s algorithm: 140¢ = 100 + 34 + 1 + 1 + 1 + 1 + 1 + 1.
・Optimal: 140¢ = 70 + 70.

A. No. It may not even lead to a feasible solution if c1 > 1: 7, 8, 9.

6
Cashier′s algorithm (for arbitrary coin denominations)

Q. Is cashier’s algorithm optimal for any set of denominations?

A. No. Consider U.S. postage: 1, 10, 21, 34, 70, 100, 350, 1225, 1500.
・Cashier’s algorithm: 140¢ = 100 + 34 + 1 + 1 + 1 + 1 + 1 + 1.
・Optimal: 140¢ = 70 + 70.

A. No. It may not even lead to a feasible solution if c1 > 1: 7, 8, 9.

・Cashier’s algorithm: 15¢ = 9 + ?.

6
Cashier′s algorithm (for arbitrary coin denominations)

Q. Is cashier’s algorithm optimal for any set of denominations?

A. No. Consider U.S. postage: 1, 10, 21, 34, 70, 100, 350, 1225, 1500.
・Cashier’s algorithm: 140¢ = 100 + 34 + 1 + 1 + 1 + 1 + 1 + 1.
・Optimal: 140¢ = 70 + 70.

A. No. It may not even lead to a feasible solution if c1 > 1: 7, 8, 9.

・Cashier’s algorithm: 15¢ = 9 + ?.
・Optimal: 15¢ = 7 + 8.
6
 Properties of any optimal solution (for U.S. coin denominations)

Property. Number of pennies ≤ 4

Pf. Replace 5 pennies with 1 nickel

Property. Number of nickels ≤ 1

Property. Number of quarters ≤ 3.

Property. Number of nickels + number of dimes ≤ 2

Pf
・Recall: ≤ 1nickel.
・Replace 3 dimes and 0 nickels with 1 quarter and 1 nickel
・Replace 2 dimes and 1 nickel with 1 quarter.

dollar quarter dimes nickels pennies

(100¢) (25¢) (10¢) (5¢) (1¢) 7
.

Optimality of cashier′s algorithm (for U.S. coin denominations)

Theorem. Cashier’s algorithm is optimal for U.S. coins { 1, 5, 10, 25, 100 }
Pf. [ by induction on amount to be paid x
・Consider optimal way to change ck ≤ x < ck+1 : greedy takes coin k
・We claim that any optimal solution must take coin k
- if not, it needs enough coins of type c1, …, ck–1 to add up to x
- table below indicates no optimal solution can do thi
・Problem reduces to coin-changing x – ck cents, which, by induction,
is optimally solved by cashier’s algorithm. ▪

all optimal solution max value of coin denomination

k ck
must satisfy c1, c2, …, ck–1 in any optimal solution

1 1 P ≤ 4 –

2 5 N ≤ 1 4

3 10 N+D ≤ 2 4+5=9

4 25 Q ≤ 3 20 + 4 = 24

5 100 no limit 75 + 24 = 99
8
s

 4. G REEDY A LGORITHMS I

‣ coin changin
‣ interval schedulin
‣ interval partitionin
‣ scheduling to minimize lateness

SECTION 4.1
g

Interval scheduling

・Job j starts at sj and nishes at fj.

h
time
0 1 2 3 4 5 6 7 8 9 10 11

10
fi
Interval scheduling

・Job j starts at sj and nishes at fj.

・Two jobs are compatible if they don’t overlap.

d
jobs d and
e
are incompatible

h
time
0 1 2 3 4 5 6 7 8 9 10 11

10
g

fi
Interval scheduling

・Job j starts at sj and nishes at fj.

・Two jobs are compatible if they don’t overlap.
・Goal: nd maximum subset of mutually compatible jobs.

d
jobs d and
e
are incompatible

h
time
0 1 2 3 4 5 6 7 8 9 10 11

10
fi
g

fi
Greedy algorithms I: quiz 2

Consider jobs in some order, taking each job provided it′s compatible
with the ones already taken. Which rule is optimal

A. [Earliest start time] Consider jobs in ascending order of sj.

B. [Earliest nish time] Consider jobs in ascending order of fj.

C. [Shortest interval] Consider jobs in ascending order of fj – sj.

D. None of the above.

11
fi
?

Greedy algorithms I: quiz 2

Consider jobs in some order, taking each job provided it′s compatible
with the ones already taken. Which rule is optimal

A. [Earliest start time] Consider jobs in ascending order of sj.

B. [Earliest nish time] Consider jobs in ascending order of fj.

C. [Shortest interval] Consider jobs in ascending order of fj – sj.

D. None of the above.

11
fi
?

Greedy algorithms I: quiz 2

Consider jobs in some order, taking each job provided it′s compatible
with the ones already taken. Which rule is optimal

A. [Earliest start time] Consider jobs in ascending order of sj.

B. [Earliest nish time] Consider jobs in ascending order of fj.

C. [Shortest interval] Consider jobs in ascending order of fj – sj.

D. None of the above.

counterexample for earliest start time

11
fi
?

Greedy algorithms I: quiz 2

Consider jobs in some order, taking each job provided it′s compatible
with the ones already taken. Which rule is optimal

A. [Earliest start time] Consider jobs in ascending order of sj.

B. [Earliest nish time] Consider jobs in ascending order of fj.

C. [Shortest interval] Consider jobs in ascending order of fj – sj.

D. None of the above.

counterexample for earliest

shorteststart time
interval

11
fi
?

Interval scheduling: earliest- nish-time- rst algorithm

EARLIEST-FINISH-TIME-FIRST (n, s1, s2, …, sn , f1, f2, …, fn)

_________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________

SORT jobs by nish times and renumber so that f1 ≤ f2 ≤ … ≤ fn.

S ← ∅. set of jobs selected

FOR j = 1 TO

IF (job j is compatible with S)

S ← S ∪ { j }.
RETURN S.
________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________

Proposition. Can implement earliest- nish-time rst in O(n log n) time.

・Keep track of job j* that was added last to S
・Job j is compatible with S iff sj ≥ fj*
・Sorting by nish times takes O(n log n) time.
12
fi
fi
n

fi
.

fi
.

fi
fi
_

Interval scheduling: analysis of earliest- nish-time- rst algorithm

Theorem. The earliest- nish-time- rst algorithm is optimal.

13
fi
fi
fi
fi
Interval scheduling: analysis of earliest- nish-time- rst algorithm

Theorem. The earliest- nish-time- rst algorithm is optimal.

Pf. [by contradiction]

13
fi
fi
fi
fi
Interval scheduling: analysis of earliest- nish-time- rst algorithm

Theorem. The earliest- nish-time- rst algorithm is optimal.

Pf. [by contradiction]

・Assume greedy is not optimal, and let’s see what happens.

13
fi
fi
fi
fi
 Interval scheduling: analysis of earliest- nish-time- rst algorithm

Theorem. The earliest- nish-time- rst algorithm is optimal.

Pf. [by contradiction]

・Assume greedy is not optimal, and let’s see what happens.
・Let i1, i2, ... ik denote set of jobs selected by greedy.

Greedy: i1 i2 ir ir+1 ... ik

13
fi
fi
fi
fi
 Interval scheduling: analysis of earliest- nish-time- rst algorithm

Theorem. The earliest- nish-time- rst algorithm is optimal.

Pf. [by contradiction]

・Assume greedy is not optimal, and let’s see what happens.
・Let i1, i2, ... ik denote set of jobs selected by greedy.
・Let j1, j2, ... jm denote set of jobs in an optimal solution with
i1 = j1, i2 = j2, ..., ir = jr for the largest possible value of r.

Greedy: i1 i2 ir ir+1 ... ik

Optimal: j1 j2 jr jr+1 ... jm

13
fi
fi
fi
fi
 Interval scheduling: analysis of earliest- nish-time- rst algorithm

Theorem. The earliest- nish-time- rst algorithm is optimal.

Pf. [by contradiction]

・Assume greedy is not optimal, and let’s see what happens.
・Let i1, i2, ... ik denote set of jobs selected by greedy.
・Let j1, j2, ... jm denote set of jobs in an optimal solution with
i1 = j1, i2 = j2, ..., ir = jr for the largest possible value of r.

Greedy: i1 i2 ir ir+1 ... ik

Optimal: j1 j2 jr jr+1 ... jm

job jr+1 exist

because m > k

13
s

fi
fi
fi
fi
 Interval scheduling: analysis of earliest- nish-time- rst algorithm

Theorem. The earliest- nish-time- rst algorithm is optimal.

Pf. [by contradiction]

・Assume greedy is not optimal, and let’s see what happens.
・Let i1, i2, ... ik denote set of jobs selected by greedy.
・Let j1, j2, ... jm denote set of jobs in an optimal solution with
i1 = j1, i2 = j2, ..., ir = jr for the largest possible value of r.

job ir+1 exists and nishes no later than jr+1

Greedy: i1 i2 ir ir+1 ... ik

Optimal: j1 j2 jr jr+1 ... jm

job jr+1 exist

because m > k

13
s

fi
fi
fi
fi
fi
 Interval scheduling: analysis of earliest- nish-time- rst algorithm

Theorem. The earliest- nish-time- rst algorithm is optimal.

Pf. [by contradiction]

・Assume greedy is not optimal, and let’s see what happens.
・Let i1, i2, ... ik denote set of jobs selected by greedy.
・Let j1, j2, ... jm denote set of jobs in an optimal solution with
i1 = j1, i2 = j2, ..., ir = jr for the largest possible value of r.

job ir+1 exists and nishes no later than jr+1

Greedy: i1 i2 ir ir+1 ... ik

Optimal: j1 j2 jr jr+1 ... jm

job jr+1 exist why not replac

because m > k job jr+1 with job ir+1?

13
s

fi
fi
fi
fi
fi
 Interval scheduling: analysis of earliest- nish-time- rst algorithm

Theorem. The earliest- nish-time- rst algorithm is optimal

Pf. [by contradiction

・Assume greedy is not optimal, and let’s see what happens
・Let i1, i2, ... ik denote set of jobs selected by greedy
・Let j1, j2, ... jm denote set of jobs in an optimal solution with
i1 = j1, i2 = j2, ..., ir = jr for the largest possible value of r.

job ir+1 exists and nishes before jr+1

Greedy: i1 i2 ir ir+1 ... ik

Optimal: j1 j2 jr ir+1 ... jm

solution still feasible and optima

(but contradicts maximality of r)

14
fi
]

fi
fi
fi
.

fi
 Greedy algorithms I: quiz 3

Suppose that each job also has a positive weight and the goal is to
nd a maximum weight subset of mutually compatible intervals.
Is the earliest- nish-time- rst algorithm still optimal?

A. Yes, because greedy algorithms are always optimal.

B. Yes, because the same proof of correctness is valid

C. No, because the same proof of correctness is no longer valid

D. No, because you could assign a huge weight to a job that overlaps
the job with the earliest nish time.

counterexample for earliest nish time

weight = 100

weight = 1

15
fi
fi
fi
fi
fi
.

 Greedy algorithms I: quiz 3

Suppose that each job also has a positive weight and the goal is to
nd a maximum weight subset of mutually compatible intervals.
Is the earliest- nish-time- rst algorithm still optimal?

A. Yes, because greedy algorithms are always optimal.

B. Yes, because the same proof of correctness is valid

C. No, because the same proof of correctness is no longer valid

D. No, because you could assign a huge weight to a job that overlaps
the job with the earliest nish time.

counterexample for earliest nish time

weight = 100

weight = 1

15
fi
fi
fi
fi
fi
.

 4. G REEDY A LGORITHMS I

‣ coin changin
‣ interval schedulin
‣ interval partitionin
‣ scheduling to minimize lateness

SECTION 4.1
g

Interval partitioning

・Lecture j starts at sj and nishes at fj

・Goal: nd minimum number of classrooms to schedule all lectures
so that no two lectures occur at the same time in the same room.

Ex. This schedule uses 4 classrooms to schedule 10 lectures.

jobs e and
are incompatible

4 e j

3 c d g

2 b h

1 a f i

9 9:30 10 10:30 11 11:30 12 12:30 1 1:30 2 2:30 3 3:30 4 4:30 time

17
fi
g

fi
.

Interval partitioning

・Lecture j starts at sj and nishes at fj

・Goal: nd minimum number of classrooms to schedule all lectures
so that no two lectures occur at the same time in the same room

Ex. This schedule uses 3 classrooms to schedule 10 lectures

intervals are ope

(need only 3 classrooms at 2pm)

3 c d f j

2 b g i

1 a e h

9 9:30 10 10:30 11 11:30 12 12:30 1 1:30 2 2:30 3 3:30 4 4:30 time

18
fi
n

fi
.

Greedy algorithms I: quiz 4

Consider lectures in some order, assigning each lecture to rst

available classroom (opening a new classroom if none is available).
Which rule is optimal

A. [Earliest start time] Consider lectures in ascending order of sj.

B. [Earliest nish time] Consider lectures in ascending order of fj.

C. [Shortest interval] Consider lectures in ascending order of fj – sj.

D. None of the above.

19
fi
?

fi
Greedy algorithms I: quiz 4

Consider lectures in some order, assigning each lecture to rst

available classroom (opening a new classroom if none is available).
Which rule is optimal

A. [Earliest start time] Consider lectures in ascending order of sj.

B. [Earliest nish time] Consider lectures in ascending order of fj.

C. [Shortest interval] Consider lectures in ascending order of fj – sj.

D. None of the above.

19
fi
?

fi
Greedy algorithms I: quiz 4

Consider lectures in some order, assigning each lecture to rst

available classroom (opening a new classroom if none is available).
Which rule is optimal

A. [Earliest start time] Consider lectures in ascending order of sj.

B. [Earliest nish time] Consider lectures in ascending order of fj.

C. [Shortest interval] Consider lectures in ascending order of fj – sj.

D. None of the above.

counterexample for earliest nish time

3
2
1
19
fi
fi
?

fi
Greedy algorithms I: quiz 4

Consider lectures in some order, assigning each lecture to rst

available classroom (opening a new classroom if none is available).
Which rule is optimal

A. [Earliest start time] Consider lectures in ascending order of sj.

B. [Earliest nish time] Consider lectures in ascending order of fj.

C. [Shortest interval] Consider lectures in ascending order of fj – sj.

D. None of the above.

counterexample for shortest interval

3
2
1
19
fi
?

fi
Interval partitioning: earliest-start-time- rst algorithm

EARLIEST-START-TIME-FIRST (n, s1, s2, …, sn , f1, f2, …, fn)

________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________

SORT lectures by start times and renumber so that s1 ≤ s2 ≤ … ≤ sn

d ← 0. number of allocated classrooms

FOR j = 1 TO
IF (lecture j is compatible with some classroom)
Schedule lecture j in any such classroom k
ELSE
Allocate a new classroom d + 1
Schedule lecture j in classroom d + 1
d ← d + 1.
RETURN schedule
________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________

20
n

fi
_

Interval partitioning: earliest-start-time- rst algorithm

Proposition. The earliest-start-time- rst algorithm can be implemented in O(n log n)

time.

Pf. Store classrooms in a priority queue (key = nish time of its last lecture).
・To determine whether lecture j is compatible with some classroom,
compare sj to key of min classroom k in priority queue
・To add lecture j to classroom k, increase key of classroom k to fj.
・Total number of priority queue operations is O(n)
・Sorting by start times takes O(n log n) time. ▪

Remark. This implementation chooses a classroom k whose nish time

of its last lecture is the earliest.

21
fi
fi
fi
.

fi
Interval partitioning: lower bound on optimal solution

Def. The depth of a set of open intervals is the maximum number of intervals that
contain any given point.

depth = 3

3 c d f j

2 b g i

1 a e h

9 9:30 10 10:30 11 11:30 12 12:30 1 1:30 2 2:30 3 3:30 4 4:30 time

22
Interval partitioning: lower bound on optimal solution

Def. The depth of a set of open intervals is the maximum number of intervals that
contain any given point.

Key observation. Number of classrooms needed ≥ depth.

depth = 3

3 c d f j

2 b g i

1 a e h

9 9:30 10 10:30 11 11:30 12 12:30 1 1:30 2 2:30 3 3:30 4 4:30 time

22
Interval partitioning: lower bound on optimal solution

Def. The depth of a set of open intervals is the maximum number of intervals that
contain any given point.

Key observation. Number of classrooms needed ≥ depth.

Q. Does minimum number of classrooms needed always equal depth?

depth = 3

3 c d f j

2 b g i

1 a e h

9 9:30 10 10:30 11 11:30 12 12:30 1 1:30 2 2:30 3 3:30 4 4:30 time

22
Interval partitioning: lower bound on optimal solution

Def. The depth of a set of open intervals is the maximum number of intervals that
contain any given point.

Key observation. Number of classrooms needed ≥ depth.

Q. Does minimum number of classrooms needed always equal depth?

A. Yes! Moreover, earliest-start-time- rst algorithm nds a schedule
whose number of classrooms equals the depth.

depth = 3

3 c d f j

2 b g i

1 a e h

9 9:30 10 10:30 11 11:30 12 12:30 1 1:30 2 2:30 3 3:30 4 4:30 time

22
fi
fi
 Interval partitioning: analysis of earliest-start-time- rst algorithm

Observation. The earliest-start-time rst algorithm never schedules two

incompatible lectures in the same classroom.

Theorem. Earliest-start-time- rst algorithm is optimal

Pf.
・Let d = number of classrooms that the algorithm allocates
・Classroom d is opened because we needed to schedule a lecture, say j,
that is incompatible with a lecture in each of d – 1 other classrooms
・Thus, these d – 1 lectures each end after sj
・Since we sorted by start time, each of these incompatible lectures start no later
than sj
・Thus, we have d lectures overlapping at time sj + ε
・Key observation ⇒ all schedules use ≥ d classrooms. ▪

23

fi
fi
.

fi
.

 4. G REEDY A LGORITHMS I

‣ coin changin
‣ interval schedulin
‣ interval partitionin
‣ scheduling to minimize lateness

SECTION 4.2
g

Scheduling to minimizing lateness

25
Scheduling to minimizing lateness

・Single resource processes one job at a time.

25
Scheduling to minimizing lateness

・Single resource processes one job at a time.

・Job j requires tj units of processing time and is due at time dj.

1 2 3 4 5 6

tj 3 2 1 4 3 2

dj 6 8 9 9 14 15

25
Scheduling to minimizing lateness

・Single resource processes one job at a time.

・Job j requires tj units of processing time and is due at time dj.
・If j starts at time sj, it nishes at time fj = sj + tj.

1 2 3 4 5 6

tj 3 2 1 4 3 2

dj 6 8 9 9 14 15

d3 = 9 d2 = 8 d6 = 15 d1 = 6 d5 = 14 d4 = 9

0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15

25
fi
Scheduling to minimizing lateness

・Single resource processes one job at a time.

・Job j requires tj units of processing time and is due at time dj.
・If j starts at time sj, it nishes at time fj = sj + tj.
・Lateness: ℓj = max { 0, fj – dj }.

1 2 3 4 5 6

tj 3 2 1 4 3 2

dj 6 8 9 9 14 15

lateness = 2 lateness = 0

d3 = 9 d2 = 8 d6 = 15 d1 = 6 d5 = 14 d4 = 9

0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15

25
fi
Scheduling to minimizing lateness

・Single resource processes one job at a time.

・Job j requires tj units of processing time and is due at time dj.
・If j starts at time sj, it nishes at time fj = sj + tj.
・Lateness: ℓj = max { 0, fj – dj }.
・Goal: schedule all jobs to minimize maximum lateness L = maxj ℓj.

1 2 3 4 5 6

tj 3 2 1 4 3 2

dj 6 8 9 9 14 15

lateness = 2 lateness = 0 max lateness = 6

d3 = 9 d2 = 8 d6 = 15 d1 = 6 d5 = 14 d4 = 9

0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15

25
fi
Greedy algorithms I: quiz 5

Schedule jobs according to some natural order. Which order minimizes

the maximum lateness

A. [shortest processing time] Ascending order of processing time tj.

B. [earliest deadline rst] Ascending order of deadline dj.

C. [smallest slack] Ascending order of slack: dj – tj.

D. None of the above.

26
fi
?

Greedy algorithms I: quiz 5

Schedule jobs according to some natural order. Which order minimizes

the maximum lateness

A. [shortest processing time] Ascending order of processing time tj.

B. [earliest deadline rst] Ascending order of deadline dj.

C. [smallest slack] Ascending order of slack: dj – tj.

D. None of the above.

26
fi
?

 Greedy algorithms I: quiz 5

Schedule jobs according to some natural order. Which order minimizes

the maximum lateness

A. [shortest processing time] Ascending order of processing time tj.

B. [earliest deadline rst] Ascending order of deadline dj.

C. [smallest slack] Ascending order of slack: dj – tj.

D. None of the above.

counterexample for shortest processing time

1 2
L =
tj 9 10
L* = 0

dj 100 10

26
9

fi
?

 Greedy algorithms I: quiz 5

Schedule jobs according to some natural order. Which order minimizes

the maximum lateness

A. [shortest processing time] Ascending order of processing time tj.

B. [earliest deadline rst] Ascending order of deadline dj.

C. [smallest slack] Ascending order of slack: dj – tj.

D. None of the above.

counterexample for smallest slack

1 2
L =
tj 1 10
L* = 1

dj 2 10

26
9

fi
?

 Minimizing lateness: earliest deadline rst

EARLIEST-DEADLINE-FIRST (n, t1, t2, …, tn , d1, d2, …, dn)

_________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________

SORT jobs by due times and renumber so that d1 ≤ d2 ≤ … ≤ dn

t←0
FOR j = 1 TO
Assign job j to interval [t, t + tj ].
sj ← t ; fj ← t + tj
t ← t + tj
RETURN intervals [s1, f1], [s2, f2], …, [sn, fn].
_________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________

max lateness L = 1

d1 = 6 d2 = 8 d3 = 9 d4 = 9 d5 = 14 d6 = 15

0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15

27
.

fi
_

 Minimizing lateness: no idle time

Observation 1. There exists an optimal schedule with no idle time.

an optimal schedule d=4 d=6 d = 12

0 1 2 3 4 5 6 7 8 9 10 11

28
 Minimizing lateness: no idle time

Observation 1. There exists an optimal schedule with no idle time.

an optimal schedule d=4 d=6 d = 12

0 1 2 3 4 5 6 7 8 9 10 11

28
 Minimizing lateness: no idle time

Observation 1. There exists an optimal schedule with no idle time.

an optimal schedule d=4 d=6 d = 12

0 1 2 3 4 5 6 7 8 9 10 11

28
 Minimizing lateness: no idle time

Observation 1. There exists an optimal schedule with no idle time.

an optimal schedule d=4 d=6 d = 12

0 1 2 3 4 5 6 7 8 9 10 11

an optimal schedul d=4 d=6 d = 12

with no idle time 0 1 2 3 4 5 6 7 8 9 10 11

28
e

 Minimizing lateness: no idle time

Observation 1. There exists an optimal schedule with no idle time.

an optimal schedule d=4 d=6 d = 12

0 1 2 3 4 5 6 7 8 9 10 11

an optimal schedul d=4 d=6 d = 12

with no idle time 0 1 2 3 4 5 6 7 8 9 10 11

Observation 2. The earliest-deadline- rst schedule has no idle time.

28
e

fi
Minimizing lateness: inversions

Def. Given a schedule S, an inversion is a pair of jobs i and j such that:

i < j but j is scheduled before i.
inversion if i < j

a schedule wit
j i
an inversion

recall: we assume the jobs are numbered so that d1 ≤ d2 ≤ … ≤ dn

29
h

Minimizing lateness: inversions

Def. Given a schedule S, an inversion is a pair of jobs i and j such that:

i < j but j is scheduled before i.
inversion if i < j

a schedule wit
j i
an inversion

recall: we assume the jobs are numbered so that d1 ≤ d2 ≤ … ≤ dn

Observation 3. The earliest-deadline- rst schedule is the unique idle-free schedule

with no inversions.

1 2 3 4 5 6 … n

29
h

fi
 Minimizing lateness: inversions

Def. Given a schedule S, an inversion is a pair of jobs i and j such that:

i < j but j is scheduled before i
inversion if i < j

a schedule wit
j i
an inversion

recall: we assume the jobs are numbered so that d1 ≤ d2 ≤ … ≤ dn

Observation 4. If an idle-free schedule has an inversion, then it has an adjacent

inversion
Pf two inverted jobs scheduled consecutively

・Suppose there are no adjacent inversions

・Let i– j be a closest inversion (which is not an adjacent inversion)
・Let k be element immediately to the right of j
・Case 1. [ j > k ] Then j–k is an adjacent inversion. (Contradiction
・Case 2. [ j < k ] Then i–k is a closer inversion since i < j < k. (Contradiction) ※

j k i
30
.

 Minimizing lateness: inversions

Def. Given a schedule S, an inversion is a pair of jobs i and j such that:

i < j but j is scheduled before i
inversion if i < j
fi

befor
exchange
j i

afte
exchange i j
f ʹj

Key claim. Exchanging two adjacent, inverted jobs i and j reduces the number of
inversions by 1 and does not increase the max lateness
Pf. Let ℓ be the lateness before the swap, and let ℓʹ be it afterwards
・ℓʹk = ℓk for all k ≠ i, j.
・ℓʹi ≤ ℓi.
・If job j is late, ℓʹj = f ʹj – dj de nition

= fi – dj j now nishes at time fi

≤ fi – di i < j ⇒ di ≤ dj
≤ ℓi . de nition
31
fi
fi
r

fi
e

 Minimizing lateness: analysis of earliest-deadline- rst algorithm

Theorem. The earliest-deadline- rst schedule S is optimal

optimal schedule ca
Pf. [by contradiction] have inversions

De ne S* to be an optimal schedule with the fewest inversions

・Can assume S* has no idle time Observation 1

・Case 1. [ S* has no inversions ] Then S = S* Observation 3

・Case 2. [ S* has an inversion

- let i–j be an adjacent inversio Observation 4

- exchanging jobs i and j decreases the number of inversions by 1

without increasing the max latenes key claim

- contradicts “fewest inversions” part of the de nition of S* ※

32
fi
n

fi
s

fi
.

fi
.

Greedy analysis strategies

Greedy algorithm stays ahead. Show that after each step of the greedy algorithm,
its solution is at least as good as any other algorithm’s.

Structural. Discover a simple “structural” bound asserting that every possible

solution must have a certain value. Then show that your algorithm always achieves
this bound.

Exchange argument. Gradually transform any solution to the one found by the
greedy algorithm without hurting its quality.

Other greedy algorithms. Gale–Shapley, Kruskal, Prim, Dijkstra, Huffman, …

33