Polyamorous Scheduling - Speaker Deck

Slide 1

Slide 1 text

Polyamorous Scheduling Sebastian Wild joint work with Leszek Gąsieniec and Ben Smith FUN with Algorithms 2024 Sebastian Wild Polyamorous Scheduling 2024-06-06 0 / 16

Slide 2

Slide 2 text

Outline 1 Introduction 1 Introduction 2 Hardness 2 Hardness 3 Approximation 3 Approximation 4 Density 4 Density 5 Conclusion 5 Conclusion Sebastian Wild Polyamorous Scheduling 2024-06-06 0 / 16

Slide 3

Slide 3 text

1 Introduction 1 Introduction Sebastian Wild Polyamorous Scheduling 2024-06-06 0 / 16

Slide 4

Slide 4 text

Polyamory Sebastian Wild Polyamorous Scheduling 2024-06-06 1 / 16

Slide 5

Slide 5 text

Polyamory Sebastian Wild Polyamorous Scheduling 2024-06-06 1 / 16

Slide 6

Slide 6 text

Polyamory Sebastian Wild Polyamorous Scheduling 2024-06-06 1 / 16

Slide 7

Slide 7 text

Poly Scheduling Given: People and pairwise Relationships ⇝ undirected graph (P, R) each relationship has a “desire growth rate” ⇝ edge weights g : R → R>0 pain of separation grows by g(e) each day Goal: perpetual schedule S : N0 → 2R that minimizes “heat”: maximal desire ever felt (minimax objective) each person can engage in one relationship per day ⇝ every day t, schedule a matching S(t) A A B B C C D D E E F F G G H H Example polycule with 8 persons: Adam, Brady, Charlie, Daisy, Eli, Frankie, Grace, and Holly Sebastian Wild Polyamorous Scheduling 2024-06-06 2 / 16

Slide 8

Slide 8 text

Poly Scheduling Given: People and pairwise Relationships ⇝ undirected graph (P, R) each relationship has a “desire growth rate” ⇝ edge weights g : R → R>0 pain of separation grows by g(e) each day Goal: perpetual schedule S : N0 → 2R that minimizes “heat”: maximal desire ever felt (minimax objective) each person can engage in one relationship per day ⇝ every day t, schedule a matching S(t) A A B B C C D D E E F F G G H H 40 ♡ 80 ♡ 16 ♡ 20 ♡ 40 ♡ 40 ♡ 40 ♡ 80 ♡ 16 ♡ 80 ♡ g(E−F) = 40 Example polycule with 8 persons: Adam, Brady, Charlie, Daisy, Eli, Frankie, Grace, and Holly Sebastian Wild Polyamorous Scheduling 2024-06-06 2 / 16

Slide 9

Slide 9 text

Poly Scheduling Given: People and pairwise Relationships ⇝ undirected graph (P, R) each relationship has a “desire growth rate” ⇝ edge weights g : R → R>0 pain of separation grows by g(e) each day Goal: perpetual schedule S : N0 → 2R that minimizes “heat”: maximal desire ever felt (minimax objective) each person can engage in one relationship per day ⇝ every day t, schedule a matching S(t) A A B B C C D D E E F F G G H H 40 ♡ 80 ♡ 16 ♡ 20 ♡ 40 ♡ 40 ♡ 40 ♡ 80 ♡ 16 ♡ 80 ♡ g(E−F) = 40 Example polycule with 8 persons: Adam, Brady, Charlie, Daisy, Eli, Frankie, Grace, and Holly Sebastian Wild Polyamorous Scheduling 2024-06-06 2 / 16

Slide 10

Slide 10 text

Poly Scheduling – Small Example Day 0 heat: 0 A♡B A♡C B♡C C♡D A A B B C C D D 40 ♡ 20 ♡ 17 ♡ 25 ♡ Sebastian Wild Polyamorous Scheduling 2024-06-06 3 / 16

Slide 11

Slide 11 text

Poly Scheduling – Small Example Day 0 heat: 40 A♡B A♡C B♡C C♡D A A B B C C D D 40 ♡ 20 ♡ 17 ♡ 25 ♡ Sebastian Wild Polyamorous Scheduling 2024-06-06 3 / 16

Slide 12

Slide 12 text

Poly Scheduling – Small Example Day 0 heat: 40 A♡B A♡C B♡C C♡D A A B B C C D D 40 ♡ 20 ♡ 17 ♡ 25 ♡ Day A♡B A♡C B♡C C♡D 0 ♥ ♥ Sebastian Wild Polyamorous Scheduling 2024-06-06 3 / 16

Slide 13

Slide 13 text

Slide 14

Slide 14 text

Poly Scheduling – Small Example Day 1 heat: 40 A♡B A♡C B♡C C♡D A A B B C C D D 40 ♡ 20 ♡ 17 ♡ 25 ♡ Day A♡B A♡C B♡C C♡D 0 ♥ ♥ Sebastian Wild Polyamorous Scheduling 2024-06-06 3 / 16

Slide 15

Slide 15 text

Poly Scheduling – Small Example Day 1 heat: 40 A♡B A♡C B♡C C♡D A A B B C C D D 40 ♡ 20 ♡ 17 ♡ 25 ♡ Day A♡B A♡C B♡C C♡D 0 ♥ ♥ 1 ♥ Sebastian Wild Polyamorous Scheduling 2024-06-06 3 / 16

Slide 16

Slide 16 text

Poly Scheduling – Small Example Day 2 heat: 80 A♡B A♡C B♡C C♡D A A B B C C D D 40 ♡ 20 ♡ 17 ♡ 25 ♡ Day A♡B A♡C B♡C C♡D 0 ♥ ♥ 1 ♥ Sebastian Wild Polyamorous Scheduling 2024-06-06 3 / 16

Slide 17

Slide 17 text

Poly Scheduling – Small Example Day 2 heat: 80 A♡B A♡C B♡C C♡D A A B B C C D D 40 ♡ 20 ♡ 17 ♡ 25 ♡ Day A♡B A♡C B♡C C♡D 0 ♥ ♥ 1 ♥ 2 ♥ ♥ Sebastian Wild Polyamorous Scheduling 2024-06-06 3 / 16

Slide 18

Slide 18 text

Poly Scheduling – Small Example Day 3 heat: 80 A♡B A♡C B♡C C♡D A A B B C C D D 40 ♡ 20 ♡ 17 ♡ 25 ♡ Day A♡B A♡C B♡C C♡D 0 ♥ ♥ 1 ♥ 2 ♥ ♥ Sebastian Wild Polyamorous Scheduling 2024-06-06 3 / 16

Slide 19

Slide 19 text

Poly Scheduling – Small Example Day 3 heat: 80 A♡B A♡C B♡C C♡D A A B B C C D D 40 ♡ 20 ♡ 17 ♡ 25 ♡ Day A♡B A♡C B♡C C♡D 0 ♥ ♥ 1 ♥ 2 ♥ ♥ 3 ♥ Sebastian Wild Polyamorous Scheduling 2024-06-06 3 / 16

Slide 20

Slide 20 text

Poly Scheduling – Small Example Day 4 heat: 80 A♡B A♡C B♡C C♡D A A B B C C D D 40 ♡ 20 ♡ 17 ♡ 25 ♡ Day A♡B A♡C B♡C C♡D 0 ♥ ♥ 1 ♥ 2 ♥ ♥ 3 ♥ Sebastian Wild Polyamorous Scheduling 2024-06-06 3 / 16

Slide 21

Slide 21 text

Poly Scheduling – Small Example Day 4 heat: 80 A♡B A♡C B♡C C♡D A A B B C C D D 40 ♡ 20 ♡ 17 ♡ 25 ♡ Day A♡B A♡C B♡C C♡D 0 ♥ ♥ 1 ♥ 2 ♥ ♥ 3 ♥ 4 ♥ ♥ Sebastian Wild Polyamorous Scheduling 2024-06-06 3 / 16

Slide 22

Slide 22 text

Poly Scheduling – Small Example Day 5 heat: 80 A♡B A♡C B♡C C♡D A A B B C C D D 40 ♡ 20 ♡ 17 ♡ 25 ♡ Day A♡B A♡C B♡C C♡D 0 ♥ ♥ 1 ♥ 2 ♥ ♥ 3 ♥ 4 ♥ ♥ Sebastian Wild Polyamorous Scheduling 2024-06-06 3 / 16

Slide 23

Slide 23 text

Poly Scheduling – Small Example Day 5 heat: 80 A♡B A♡C B♡C C♡D A A B B C C D D 40 ♡ 20 ♡ 17 ♡ 25 ♡ Day A♡B A♡C B♡C C♡D 0 ♥ ♥ 1 ♥ 2 ♥ ♥ 3 ♥ 4 ♥ ♥ 5 ♥ Sebastian Wild Polyamorous Scheduling 2024-06-06 3 / 16

Slide 24

Slide 24 text

Poly Scheduling – Small Example Day 6 heat: 80 A♡B A♡C B♡C C♡D A A B B C C D D 40 ♡ 20 ♡ 17 ♡ 25 ♡ Day A♡B A♡C B♡C C♡D 0 ♥ ♥ 1 ♥ 2 ♥ ♥ 3 ♥ 4 ♥ ♥ 5 ♥ Sebastian Wild Polyamorous Scheduling 2024-06-06 3 / 16

Slide 25

Slide 25 text

Poly Scheduling – Small Example Day 6 heat: 80 A♡B A♡C B♡C C♡D A A B B C C D D 40 ♡ 20 ♡ 17 ♡ 25 ♡ Day A♡B A♡C B♡C C♡D 0 ♥ ♥ 1 ♥ 2 ♥ ♥ 3 ♥ 4 ♥ ♥ 5 ♥ 6 ♥ ♥ Sebastian Wild Polyamorous Scheduling 2024-06-06 3 / 16

Slide 26

Slide 26 text

Poly Scheduling – Small Example Day 7 heat: 80 A♡B A♡C B♡C C♡D A A B B C C D D 40 ♡ 20 ♡ 17 ♡ 25 ♡ Day A♡B A♡C B♡C C♡D 0 ♥ ♥ 1 ♥ 2 ♥ ♥ 3 ♥ 4 ♥ ♥ 5 ♥ 6 ♥ ♥ Sebastian Wild Polyamorous Scheduling 2024-06-06 3 / 16

Slide 27

Slide 27 text

Poly Scheduling – Small Example Day 7 heat: 80 A♡B A♡C B♡C C♡D A A B B C C D D 40 ♡ 20 ♡ 17 ♡ 25 ♡ Day A♡B A♡C B♡C C♡D 0 ♥ ♥ 1 ♥ 2 ♥ ♥ 3 ♥ 4 ♥ ♥ 5 ♥ 6 ♥ ♥ 7 ♥ Sebastian Wild Polyamorous Scheduling 2024-06-06 3 / 16

Slide 28

Slide 28 text

Poly Scheduling – Small Example Day 7 heat: 80 A♡B A♡C B♡C C♡D A A B B C C D D 40 ♡ 20 ♡ 17 ♡ 25 ♡ Day A♡B A♡C B♡C C♡D 0 ♥ ♥ 1 ♥ 2 ♥ ♥ 3 ♥ 4 ♥ ♥ 5 ♥ 6 ♥ ♥ 7 ♥ S∗ = {AB, CD}, {AC}, {AB, CD}, {BC} with heat h(S) = 80 Observation: schedules without follows from finite state space loss of generality periodic ∃T ∀t S(t) = S(t + T) Sebastian Wild Polyamorous Scheduling 2024-06-06 3 / 16

Slide 29

Slide 29 text

Decision problem Optimization input: A A B B C C D D 40 ♡ 20 ♡ 17 ♡ 25 ♡ Optimization Poly Scheduling Sebastian Wild Polyamorous Scheduling 2024-06-06 4 / 16

Slide 30

Slide 30 text

Decision problem Optimization input: A A B B C C D D 40 ♡ 20 ♡ 17 ♡ 25 ♡ Optimization Poly Scheduling Maximal frequencies to remain below heat 80: heat: 80 A♡B A♡C B♡C C♡D 2 4 4 3 frequency = heat growth rate Decision instance: A A B B C C D D 2 ♡ 4 ♡ 4 ♡ 3 ♡ A and C must meet (at least) once in every 4-day period Decision Poly Scheduling Sebastian Wild Polyamorous Scheduling 2024-06-06 4 / 16

Slide 31

Slide 31 text

Decision problem Optimization input: A A B B C C D D 40 ♡ 20 ♡ 17 ♡ 25 ♡ Optimization Poly Scheduling Maximal frequencies to remain below heat 100: heat: 100 A♡B A♡C B♡C C♡D 2 5 5 4 frequency = heat growth rate Decision instance: A A B B C C D D 2 ♡ 5 ♡ 5 ♡ 4 ♡ A and C must meet (at least) once in every 5-day period Decision Poly Scheduling Sebastian Wild Polyamorous Scheduling 2024-06-06 4 / 16

Slide 32

Slide 32 text

Decision problem Optimization input: A A B B C C D D 40 ♡ 20 ♡ 17 ♡ 25 ♡ Optimization Poly Scheduling Maximal frequencies to remain below heat 79.99: heat: 79.99 A♡B A♡C B♡C C♡D 1 3 4 3 frequency = heat growth rate Decision instance: A A B B C C D D 1 ♡ 3 ♡ 4 ♡ 3 ♡ Decision Poly Scheduling Sebastian Wild Polyamorous Scheduling 2024-06-06 4 / 16

Slide 33

Slide 33 text

Slide 34

Slide 34 text

Decision problem Optimization input: A A B B C C D D 40 ♡ 20 ♡ 17 ♡ 25 ♡ Optimization Poly Scheduling Maximal frequencies to remain below heat 79.99: heat: 79.99 A♡B A♡C B♡C C♡D 1 3 4 3 frequency = heat growth rate Decision instance: A A B B C C D D 1 ♡ 3 ♡ 4 ♡ 3 ♡ Decision Poly Scheduling For this instance: schedule S∗ has heat 80 S∗ = {AB, CD}, {AC}, {AB, CD}, {BC} heat 80 − ε leaves A unschedulable ⇝ optimal heat h∗ = 80 Sebastian Wild Polyamorous Scheduling 2024-06-06 4 / 16

Slide 35

Slide 35 text

Decision problem Optimization input: A A B B C C D D 40 ♡ 20 ♡ 17 ♡ 25 ♡ Optimization Poly Scheduling Maximal frequencies to remain below heat 79.99: heat: 79.99 A♡B A♡C B♡C C♡D 1 3 4 3 frequency = heat growth rate Decision instance: A A B B C C D D 1 ♡ 3 ♡ 4 ♡ 3 ♡ Decision Poly Scheduling For this instance: schedule S∗ has heat 80 S∗ = {AB, CD}, {AC}, {AB, CD}, {BC} heat 80 − ε leaves A unschedulable ⇝ optimal heat h∗ = 80 Sebastian Wild Polyamorous Scheduling 2024-06-06 4 / 16

Slide 36

Slide 36 text

Outline 1 Introduction 1 Introduction 2 Hardness 2 Hardness 3 Approximation 3 Approximation 4 Density 4 Density 5 Conclusion 5 Conclusion Sebastian Wild Polyamorous Scheduling 2024-06-06 4 / 16

Slide 37

Slide 37 text

2 Hardness 2 Hardness Sebastian Wild Polyamorous Scheduling 2024-06-06 4 / 16

Slide 38

Slide 38 text

Unweighted case Do poly people just make their life complicated – or are they facing a computationally hard problem? Let’s consider a simple special case: Non-hierarchical Poly Scheduling (optimization): all edges have weight 1 (decision): all edges have the same frequency f ⇝ feasible schedule for decision problem ⇐⇒ schedule of heat ⩽ f in optimization problem ⇝ Equivalent to Edge minimal #colors χ1 to color all edges s.t. no 2 incident to same vertex have same color Coloring (a.k.a. Chromatic Index) A A B B C C D D E E 1 ♡ 1 ♡ 1 ♡ 1 ♡ 1 ♡ 1 ♡ 1 ♡ 1 ♡ Edge Coloring =⇒ Poly Schedule: Schedule = Round Robin of χ1 color classes heat of schedule χ1 . Edge Coloring ⇐= Poly Schedule: Color = first day edge is scheduled Heat h ⇝ at most h colors Sebastian Wild Polyamorous Scheduling 2024-06-06 5 / 16

Slide 39

Slide 39 text

Slide 40

Slide 40 text

Slide 41

Slide 41 text

Slide 42

Slide 42 text

Slide 43

Slide 43 text

Slide 44

Slide 44 text

Slide 45

Slide 45 text

Slide 46

Slide 46 text

Slide 47

Slide 47 text

Chromatic Index Problem Recall: χ1 = chromatic index of graph G = #colors in proper edge coloring ⇝ Obivous: χ1 ⩾ ∆ maximum degree in G Vizing’s Theorem (1965): χ1 ⩽ ∆ + 1 1 ∆ + 1 coloring can always be found in efficiently. running time very recently improved to ˜ O(mn1/3) (arxiv:2405.15449) 2 Computing χ1 exactly (∆ or ∆ + 1!) is NP-hard even for 3-regular graphs! (Holyer 1981) Implications for (Non-hierarchical) Poly Scheduling 1 Decision Poly Scheduling is NP-hard. 2 Optimization Poly Scheduling does not admit poly-time (1 + δ)-approx for δ < 1/3 unless P = NP. ⇝ such an approximation would decide ∆ vs ∆ + 1 for ∆ = 3. Sebastian Wild Polyamorous Scheduling 2024-06-06 6 / 16

Slide 48

Slide 48 text

Slide 49

Slide 49 text

Slide 50

Slide 50 text

Slide 51

Slide 51 text

Bonus: SAT-Reduction Alternative hardness of approximation: Reduction from (decision version of) MAX-3SAT Given formula φ over m clauses and n variables, can ⩾ k clauses be satisfied simultaneously? Key idea: relation scheduled on odd (even) days = True (False) doesn’t quite work, but with slots module 6 it does. Sebastian Wild Polyamorous Scheduling 2024-06-06 7 / 16

Slide 52

Slide 52 text

Outline 1 Introduction 1 Introduction 2 Hardness 2 Hardness 3 Approximation 3 Approximation 4 Density 4 Density 5 Conclusion 5 Conclusion Sebastian Wild Polyamorous Scheduling 2024-06-06 7 / 16

Slide 53

Slide 53 text

3 Approximation 3 Approximation Sebastian Wild Polyamorous Scheduling 2024-06-06 7 / 16

Slide 54

Slide 54 text

Edge-Coloring Approximation Recall: Can compute (∆ + 1)-coloring efficiently ⇝ ∆ + 1 ∆ -approx for unweighted instances any valid schedule must reach heat h ⩾ ∆ on a degree-∆ vertex ⇝ in general: ∆ + 1 ∆ · gmax gmin -approx and (simultaneously) (∆ + 1) could be Ω(n) ... -approx unbounded any schedule reaches heat h ⩾ ∆ · gmin on a degree-∆ vertex any schedule reaches heat h ⩾ gmax Colors Round Robin schedule achieves h ⩽ (∆ + 1)gmax Sebastian Wild Polyamorous Scheduling 2024-06-06 8 / 16

Slide 55

Slide 55 text

Slide 56

Slide 56 text

Slide 57

Slide 57 text

Slide 58

Slide 58 text

Slide 59

Slide 59 text

Layering approximation Partition edges into exponential “layers” by growth rate Layers i = 0, . . . L − 1 contain edges with gmax 2i+1 ⩽ g(e) ⩽ gmax 2i Layer L the rest, i.e., g(e) ⩽ gmax 2L ⇝ within one layer gmax /gmin ⩽ 2 ⇝ Color Round Robin within layer gives good approximation. Also schedule layers Round Robin ⇝ Heat of overall schedule ⩽ h with ∆i = max degree in layer i h = maxi∈[0..L] (L + 1) · (∆i + 1) gmax 2i ⇝ each layer (in isolation) induces lower bound, so h∗ ⩾ h with h = max maxi∈[0..L] ∆i · gmax 2i + 1 , gmax growth rate A B C D E F G H A B C D E F G H A B C D E F G H A B C D E F G H gmax gmax /2 gmax /22 gmax /23 gmax /24 Sebastian Wild Polyamorous Scheduling 2024-06-06 9 / 16

Slide 60

Slide 60 text

Slide 61

Slide 61 text

Slide 62

Slide 62 text

Slide 63

Slide 63 text

Layering approximation [2] Distinguish cases where max in h is attained. 1 Layer i < L dominates: h = (L + 1)(∆i ) gmax 2i we by definition: h ⩾ ∆i · gmax 2i+1 ⇝ 3(L + 1)-approximation (for ∆i ⩾ 2; otherwise trivial) 2 Layer L dominates (a.k.a. otherwise) ⇝ h = (L + 1)(∆L + 1) gmax 2L by definition: h ⩾ gmax ⇝ (L + 1)(∆L + 1)2−L-approximation Equating both cases suggests L = log2 (∆L + 1) − O(1) ⇝ 3⌈log2 (∆ + 1)⌉-approximation Recall: h = max i∈[0..L] (L + 1) · (∆i + 1) gmax 2i h = max max i∈[0..L] ∆i · gmax 2i+1 , gmax Sebastian Wild Polyamorous Scheduling 2024-06-06 10 / 16

Slide 64

Slide 64 text

Slide 65

Slide 65 text

Slide 66

Slide 66 text

Slide 67

Slide 67 text

Outline 1 Introduction 1 Introduction 2 Hardness 2 Hardness 3 Approximation 3 Approximation 4 Density 4 Density 5 Conclusion 5 Conclusion Sebastian Wild Polyamorous Scheduling 2024-06-06 10 / 16

Slide 68

Slide 68 text

4 Density 4 Density Sebastian Wild Polyamorous Scheduling 2024-06-06 10 / 16

Slide 69

Slide 69 text

Special Case: Pinwheel Scheduling The Pinwheel Scheduling Problem (Holte et.al., 1989): Given k tasks (each with frequency ai ∈ N) Is there a perpetual schedule for a (single) agent who can perform one task per day such that every section of length ai contains at least one instance of task i? Example: A = [3, 4, 5, 8] k=4 Task 1 2 3 4 Frequency 3 4 5 8 S = 1, 2, 4, 1, 3, 2, 1, 3 3 4 8 3 5 4 3 5 Sebastian Wild Polyamorous Scheduling 2024-06-06 11 / 16

Slide 70

Slide 70 text

Slide 71

Slide 71 text

Special Case: Pinwheel Scheduling The Pinwheel Scheduling Problem (Holte et.al., 1989): Given k tasks (each with frequency ai ∈ N) Is there a perpetual schedule for a (single) agent who can perform one task per day such that every section of length ai contains at least one instance of task i? Example: A = [3, 4, 5, 8] k=4 Task 1 2 3 4 Frequency 3 4 5 8 S = 1, 2, 4, 1, 3, 2, 1, 3 3 4 8 3 5 4 3 5 Decision version of Bamboo Garden Trimming (Bilò et al., FUN 2020/21) Sebastian Wild Polyamorous Scheduling 2024-06-06 11 / 16

Slide 72

Slide 72 text

Slide 73

Slide 73 text

Slide 74

Slide 74 text

Slide 75

Slide 75 text

Slide 76

Slide 76 text

Slide 77

Slide 77 text

Slide 78

Slide 78 text

Slide 79

Slide 79 text

Slide 80

Slide 80 text

The 5/6 Density Conjecture Density d(A) = k i=1 1/ai [2, 3, M] 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 ... 5/6 Density Conjecture (Chan and Chin, 1993): All Pinwheel scheduling instances where d ⩽ 5/6 are schedulable. Conjecture is proven ⇝ upcoming STOC (Akitoshi Kawamura, 2024) Theorem Sebastian Wild Polyamorous Scheduling 2024-06-06 12 / 16

Slide 81

Slide 81 text

The 5/6 Density Conjecture Density d(A) = k i=1 1/ai [2, 3, M] 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 ... 5/6 Density Conjecture (Chan and Chin, 1993): All Pinwheel scheduling instances where d ⩽ 5/6 are schedulable. Conjecture is proven ⇝ upcoming STOC (Akitoshi Kawamura, 2024) Theorem ⇝ Similar result possible for Poly Scheduling? Sebastian Wild Polyamorous Scheduling 2024-06-06 12 / 16

Slide 82

Slide 82 text

Fractional Scheduling lower-bound tool: fractional relaxation ⇝ we can spend a fraction yM of each day on matching M ⇝ without loss of generality: all days the same! ⇝ Best fractional solution given by LP: M = set of maximal matchings min ¯ h s. t. M∈M yM ⩽ 1 1 M∈M:e∈M yM · ge ⩽ ¯ h ∀e ∈ R yM ∈ [0, 1] ∀M ∈ M Sebastian Wild Polyamorous Scheduling 2024-06-06 13 / 16

Slide 83

Slide 83 text

Slide 84

Slide 84 text

Slide 85

Slide 85 text

Slide 86

Slide 86 text

Fractional Scheduling lower-bound tool: fractional relaxation ⇝ we can spend a fraction yM of each day on matching M ⇝ without loss of generality: all days the same! ⇝ Best fractional solution given by LP: after change of variable ¯ h = 1 ℓ and slight massaging M = set of maximal matchings min ¯ h s. t. M∈M yM ⩽ 1 1 M∈M:e∈M yM · ge ⩽ ¯ h ∀e ∈ R yM ∈ [0, 1] ∀M ∈ M 1 ℓ 1 ℓ Sebastian Wild Polyamorous Scheduling 2024-06-06 13 / 16

Slide 87

Slide 87 text

Fractional Scheduling lower-bound tool: fractional relaxation ⇝ we can spend a fraction yM of each day on matching M ⇝ without loss of generality: all days the same! ⇝ Best fractional solution given by LP: after change of variable ¯ h = 1 ℓ and slight massaging M = set of maximal matchings min ¯ h s. t. M∈M yM ⩽ 1 1 M∈M:e∈M yM · ge ⩽ ¯ h ∀e ∈ R yM ∈ [0, 1] ∀M ∈ M ℓ M∈M:e∈M yM 1 ge ⩾ 1 ℓ Sebastian Wild Polyamorous Scheduling 2024-06-06 13 / 16

Slide 88

Slide 88 text

Fractional Scheduling lower-bound tool: fractional relaxation ⇝ we can spend a fraction yM of each day on matching M ⇝ without loss of generality: all days the same! ⇝ Best fractional solution given by LP: after change of variable ¯ h = 1 ℓ and slight massaging M = set of maximal matchings min ¯ h s. t. M∈M yM ⩽ 1 1 M∈M:e∈M yM · ge ⩽ ¯ h ∀e ∈ R yM ∈ [0, 1] ∀M ∈ M ℓ M∈M:e∈M yM 1 ge ⩾ ℓ max Sebastian Wild Polyamorous Scheduling 2024-06-06 13 / 16

Slide 89

Slide 89 text

Dual LP → Density Fractional Poly Scheduling LP max ℓ s. t. M∈M yM ⩽ 1 1 ge M∈M:e∈M yM ⩾ ℓ ∀e ∈ R yM ⩾ 0 ∀M ∈ M LP has exponentially many variables ... but few constraints ⇝ Consider the dual LP: min x s. t. e∈R ze ⩾ 1 e∈M ze ge ⩽ x ∀M ∈ M ze ⩾ 0 ∀e ∈ R ⇝ any feasible solution (x,⃗ z) lower bounds optimal heat h∗! Theorem (Fractional lower bound) For any ze ∈ [0, 1], for e ∈ R, with e∈R ze = 1, we have h∗ ⩾ ¯ h(z) = 1 maxM∈M e∈M ze ge . Sebastian Wild Polyamorous Scheduling 2024-06-06 14 / 16

Slide 90

Slide 90 text

Outline 1 Introduction 1 Introduction 2 Hardness 2 Hardness 3 Approximation 3 Approximation 4 Density 4 Density 5 Conclusion 5 Conclusion Sebastian Wild Polyamorous Scheduling 2024-06-06 14 / 16

Slide 91

Slide 91 text

5 Conclusion 5 Conclusion Sebastian Wild Polyamorous Scheduling 2024-06-06 14 / 16

Slide 92

Slide 92 text

Conclusion What you’ve seen: Poly Scheduling is a highly desired practical problem but it’s hard NP-hard hard to approximate better than 4/3 best known polytime approximation only O(log ∆) = O(log n) Some questions: 1 Best possible approximation ratio? 2 Better results for restricted versions (e.g., special types of graphs) 3 Density threshold that guarantees schedule? Sebastian Wild Polyamorous Scheduling 2024-06-06 15 / 16

Slide 93

Slide 93 text

Sebastian Wild Polyamorous Scheduling 2024-06-06 16 / 16

Slide 94

Slide 94 text

Icons made by Freepik, Gregor Cresnar, Those Icons, Smashicons, Good Ware, Pause08, and Madebyoliver from www.flaticon.com. Vector graphics from Pressfoto, brgfx, macrovector and Jannoon028 on freepik.com Other photos from www.pixabay.com. Sebastian Wild Polyamorous Scheduling 2024-06-06 17 / 16