75
Joint work with Yuval Peres, Mikkel Thorup, Peter Winkler and Uri Zwick Overhang Bounds Mike Paterson DIMAP & Dept of Computer Science University of Warwic QuickTime™ a TIFF (LZW) dec are needed to se

Joint work with Yuval Peres, Mikkel Thorup, Peter Winkler and Uri Zwick Overhang Bounds Mike Paterson DIMAP & Dept of Computer Science University of Warwick

Embed Size (px)

Citation preview

Page 1: Joint work with Yuval Peres, Mikkel Thorup, Peter Winkler and Uri Zwick Overhang Bounds Mike Paterson DIMAP & Dept of Computer Science University of Warwick

Joint work with Yuval Peres, Mikkel Thorup, Peter Winkler and Uri Zwick

Overhang Bounds

Mike Paterson

DIMAP & Dept of Computer Science

University of Warwick

QuickTime™ and aTIFF (LZW) decompressor

are needed to see this picture.

Page 2: Joint work with Yuval Peres, Mikkel Thorup, Peter Winkler and Uri Zwick Overhang Bounds Mike Paterson DIMAP & Dept of Computer Science University of Warwick

The classical solution

Harmonic Stacks

Using n blocks we can get an overhang of

Page 3: Joint work with Yuval Peres, Mikkel Thorup, Peter Winkler and Uri Zwick Overhang Bounds Mike Paterson DIMAP & Dept of Computer Science University of Warwick

Diamonds

The 4-diamond is balanced

Page 4: Joint work with Yuval Peres, Mikkel Thorup, Peter Winkler and Uri Zwick Overhang Bounds Mike Paterson DIMAP & Dept of Computer Science University of Warwick

Diamonds

The 5-diamond is …

Page 5: Joint work with Yuval Peres, Mikkel Thorup, Peter Winkler and Uri Zwick Overhang Bounds Mike Paterson DIMAP & Dept of Computer Science University of Warwick

Diamonds?

… unbalanced!

Page 6: Joint work with Yuval Peres, Mikkel Thorup, Peter Winkler and Uri Zwick Overhang Bounds Mike Paterson DIMAP & Dept of Computer Science University of Warwick

What really happens?

Page 7: Joint work with Yuval Peres, Mikkel Thorup, Peter Winkler and Uri Zwick Overhang Bounds Mike Paterson DIMAP & Dept of Computer Science University of Warwick

What really happens!

Page 8: Joint work with Yuval Peres, Mikkel Thorup, Peter Winkler and Uri Zwick Overhang Bounds Mike Paterson DIMAP & Dept of Computer Science University of Warwick

Small optimal stacks

Overhang = 1.16789Blocks = 4

Overhang = 1.30455Blocks = 5

Overhang = 1.4367Blocks = 6

Overhang = 1.53005Blocks = 7

Page 9: Joint work with Yuval Peres, Mikkel Thorup, Peter Winkler and Uri Zwick Overhang Bounds Mike Paterson DIMAP & Dept of Computer Science University of Warwick

Small optimal stacks

Overhang = 2.14384Blocks = 16

Overhang = 2.1909Blocks = 17

Overhang = 2.23457Blocks = 18

Overhang = 2.27713Blocks = 19

Note “spinality”

Page 10: Joint work with Yuval Peres, Mikkel Thorup, Peter Winkler and Uri Zwick Overhang Bounds Mike Paterson DIMAP & Dept of Computer Science University of Warwick

Support and balancing blocks

Principalblock

Support set

Balancing

set

Page 11: Joint work with Yuval Peres, Mikkel Thorup, Peter Winkler and Uri Zwick Overhang Bounds Mike Paterson DIMAP & Dept of Computer Science University of Warwick

Support and balancing blocks

Principalblock

Support set

Balancing

set

Page 12: Joint work with Yuval Peres, Mikkel Thorup, Peter Winkler and Uri Zwick Overhang Bounds Mike Paterson DIMAP & Dept of Computer Science University of Warwick

Principalblock

Support set

Stacks with downward external

forces acting on them

Loaded stacks

Size =

number of blocks

+ sum of external

forces.

Page 13: Joint work with Yuval Peres, Mikkel Thorup, Peter Winkler and Uri Zwick Overhang Bounds Mike Paterson DIMAP & Dept of Computer Science University of Warwick

Principalblock

Support set

Stacks in which the support set contains

only one block at each level

Spinal stacks

Assumed to be optimal in:

J.F. Hall, Fun with stacking Blocks, American Journal of Physics 73(12), 1107-1116, 2005.

Page 14: Joint work with Yuval Peres, Mikkel Thorup, Peter Winkler and Uri Zwick Overhang Bounds Mike Paterson DIMAP & Dept of Computer Science University of Warwick

Optimal spinal stacks

Optimality condition:

Page 15: Joint work with Yuval Peres, Mikkel Thorup, Peter Winkler and Uri Zwick Overhang Bounds Mike Paterson DIMAP & Dept of Computer Science University of Warwick

Spinal overhang

Let S (n) be the maximal overhang achievable using a spinal stack with n blocks.

Let S*(n) be the maximal overhang achievable using a loaded spinal stack on total weight n.

Theorem:

A factor of 2 improvement over harmonic stacks!

Conjecture:

Page 16: Joint work with Yuval Peres, Mikkel Thorup, Peter Winkler and Uri Zwick Overhang Bounds Mike Paterson DIMAP & Dept of Computer Science University of Warwick

Optimal 100-block spinal stack

Spine

Shield

Towers

Page 17: Joint work with Yuval Peres, Mikkel Thorup, Peter Winkler and Uri Zwick Overhang Bounds Mike Paterson DIMAP & Dept of Computer Science University of Warwick

Optimal weight 100 loaded spinal stack

Page 18: Joint work with Yuval Peres, Mikkel Thorup, Peter Winkler and Uri Zwick Overhang Bounds Mike Paterson DIMAP & Dept of Computer Science University of Warwick

Loaded spinal stack + shield

Page 19: Joint work with Yuval Peres, Mikkel Thorup, Peter Winkler and Uri Zwick Overhang Bounds Mike Paterson DIMAP & Dept of Computer Science University of Warwick

spinal stack + shield + towers

Page 20: Joint work with Yuval Peres, Mikkel Thorup, Peter Winkler and Uri Zwick Overhang Bounds Mike Paterson DIMAP & Dept of Computer Science University of Warwick

Are spinal stacks optimal?

No!

Support set is not spinal!

Overhang = 2.32014Blocks = 20

Tiny gap

Page 21: Joint work with Yuval Peres, Mikkel Thorup, Peter Winkler and Uri Zwick Overhang Bounds Mike Paterson DIMAP & Dept of Computer Science University of Warwick

Optimal 30-block stack

Overhang = 2.70909Blocks = 30

Page 22: Joint work with Yuval Peres, Mikkel Thorup, Peter Winkler and Uri Zwick Overhang Bounds Mike Paterson DIMAP & Dept of Computer Science University of Warwick

Optimal (?) weight 100 construction

Overhang = 4.2390Blocks = 49

Weight = 100

Page 23: Joint work with Yuval Peres, Mikkel Thorup, Peter Winkler and Uri Zwick Overhang Bounds Mike Paterson DIMAP & Dept of Computer Science University of Warwick

“Parabolic” constructions

6-stack

Number of blocks in d-stack: Overhang:Balanced!

Page 24: Joint work with Yuval Peres, Mikkel Thorup, Peter Winkler and Uri Zwick Overhang Bounds Mike Paterson DIMAP & Dept of Computer Science University of Warwick

“Parabolic” constructions

6-slab

5-slab

4-slab

Page 25: Joint work with Yuval Peres, Mikkel Thorup, Peter Winkler and Uri Zwick Overhang Bounds Mike Paterson DIMAP & Dept of Computer Science University of Warwick

r-slab

r-slab

Page 26: Joint work with Yuval Peres, Mikkel Thorup, Peter Winkler and Uri Zwick Overhang Bounds Mike Paterson DIMAP & Dept of Computer Science University of Warwick

(r -1) - slab within an r - slab

(r-1)-slab

Nested inductions

Page 27: Joint work with Yuval Peres, Mikkel Thorup, Peter Winkler and Uri Zwick Overhang Bounds Mike Paterson DIMAP & Dept of Computer Science University of Warwick
Page 28: Joint work with Yuval Peres, Mikkel Thorup, Peter Winkler and Uri Zwick Overhang Bounds Mike Paterson DIMAP & Dept of Computer Science University of Warwick

“Smooth” parabola?

QuickTime™ and aTIFF (LZW) decompressor

are needed to see this picture.

Stacks with monotonic right contour can achieve only about ln n overhang [theorem above]

No good!

Page 29: Joint work with Yuval Peres, Mikkel Thorup, Peter Winkler and Uri Zwick Overhang Bounds Mike Paterson DIMAP & Dept of Computer Science University of Warwick

“Vases”

Weight = 1151.76

Blocks = 1043

Overhang = 10

Page 30: Joint work with Yuval Peres, Mikkel Thorup, Peter Winkler and Uri Zwick Overhang Bounds Mike Paterson DIMAP & Dept of Computer Science University of Warwick

“Oil lamps”

Weight = 1112.84

Blocks = 921

Overhang = 10

Page 31: Joint work with Yuval Peres, Mikkel Thorup, Peter Winkler and Uri Zwick Overhang Bounds Mike Paterson DIMAP & Dept of Computer Science University of Warwick

What about an upper bound?

Ωnis a lower bound

for overhang with n blocks

Can we do better?

Page 32: Joint work with Yuval Peres, Mikkel Thorup, Peter Winkler and Uri Zwick Overhang Bounds Mike Paterson DIMAP & Dept of Computer Science University of Warwick

Equilibrium

F1 + F2 + F3 = F4 + F5

x1 F1+ x2 F2+ x3 F3 = x4 F4+ x5 F5

Force equation

Moment equation

F1

F5F4

F3

F2

Page 33: Joint work with Yuval Peres, Mikkel Thorup, Peter Winkler and Uri Zwick Overhang Bounds Mike Paterson DIMAP & Dept of Computer Science University of Warwick

Forces between blocks

Assumption: No friction.All forces are vertical.

Equivalent sets of forces

Page 34: Joint work with Yuval Peres, Mikkel Thorup, Peter Winkler and Uri Zwick Overhang Bounds Mike Paterson DIMAP & Dept of Computer Science University of Warwick
Page 35: Joint work with Yuval Peres, Mikkel Thorup, Peter Winkler and Uri Zwick Overhang Bounds Mike Paterson DIMAP & Dept of Computer Science University of Warwick
Page 36: Joint work with Yuval Peres, Mikkel Thorup, Peter Winkler and Uri Zwick Overhang Bounds Mike Paterson DIMAP & Dept of Computer Science University of Warwick
Page 37: Joint work with Yuval Peres, Mikkel Thorup, Peter Winkler and Uri Zwick Overhang Bounds Mike Paterson DIMAP & Dept of Computer Science University of Warwick
Page 38: Joint work with Yuval Peres, Mikkel Thorup, Peter Winkler and Uri Zwick Overhang Bounds Mike Paterson DIMAP & Dept of Computer Science University of Warwick
Page 39: Joint work with Yuval Peres, Mikkel Thorup, Peter Winkler and Uri Zwick Overhang Bounds Mike Paterson DIMAP & Dept of Computer Science University of Warwick
Page 40: Joint work with Yuval Peres, Mikkel Thorup, Peter Winkler and Uri Zwick Overhang Bounds Mike Paterson DIMAP & Dept of Computer Science University of Warwick

Distributions

Page 41: Joint work with Yuval Peres, Mikkel Thorup, Peter Winkler and Uri Zwick Overhang Bounds Mike Paterson DIMAP & Dept of Computer Science University of Warwick

Moments and spread

j-th moment

Center of mass

Spread

NB important measure

Page 42: Joint work with Yuval Peres, Mikkel Thorup, Peter Winkler and Uri Zwick Overhang Bounds Mike Paterson DIMAP & Dept of Computer Science University of Warwick

Signed distributions

Page 43: Joint work with Yuval Peres, Mikkel Thorup, Peter Winkler and Uri Zwick Overhang Bounds Mike Paterson DIMAP & Dept of Computer Science University of Warwick

MovesA move is a signed distribution

with M0[ ] = M1[ ] = 0 whose support

is contained in an interval of length 1

A move is applied by adding it to a distribution.

A move can be applied only if the resulting signed distribution is a distribution.

Page 44: Joint work with Yuval Peres, Mikkel Thorup, Peter Winkler and Uri Zwick Overhang Bounds Mike Paterson DIMAP & Dept of Computer Science University of Warwick

Equilibrium

F1 + F2 + F3 = F4 + F5

x1 F1+ x2 F2+ x3 F3 = x4 F4+ x5 F5

Force equation

Moment equation

F1

F5F4

F3

F2

Recall!

Page 45: Joint work with Yuval Peres, Mikkel Thorup, Peter Winkler and Uri Zwick Overhang Bounds Mike Paterson DIMAP & Dept of Computer Science University of Warwick

MovesA move is a signed distribution

with M0[ ] = M1[ ] = 0 whose support

is contained in an interval of length 1

A move is applied by adding it to a distribution.

A move can be applied only if the resulting signed distribution is a distribution.

Page 46: Joint work with Yuval Peres, Mikkel Thorup, Peter Winkler and Uri Zwick Overhang Bounds Mike Paterson DIMAP & Dept of Computer Science University of Warwick

Move sequences

Page 47: Joint work with Yuval Peres, Mikkel Thorup, Peter Winkler and Uri Zwick Overhang Bounds Mike Paterson DIMAP & Dept of Computer Science University of Warwick

Extreme moves

Moves all the mass within the interval to the endpoints

Page 48: Joint work with Yuval Peres, Mikkel Thorup, Peter Winkler and Uri Zwick Overhang Bounds Mike Paterson DIMAP & Dept of Computer Science University of Warwick

Lossy moves

If is a move in [c-½,c+½] then

A lossy move removes one unit of mass from position c

Alternatively, a lossy move freezes one unit of mass at position c

Page 49: Joint work with Yuval Peres, Mikkel Thorup, Peter Winkler and Uri Zwick Overhang Bounds Mike Paterson DIMAP & Dept of Computer Science University of Warwick

Overhang and mass movementIf there is an n-block stack that achieves an overhang of d, then

n–1 lossy moves

Page 50: Joint work with Yuval Peres, Mikkel Thorup, Peter Winkler and Uri Zwick Overhang Bounds Mike Paterson DIMAP & Dept of Computer Science University of Warwick

Main theorem

Page 51: Joint work with Yuval Peres, Mikkel Thorup, Peter Winkler and Uri Zwick Overhang Bounds Mike Paterson DIMAP & Dept of Computer Science University of Warwick

Four stepsShift half mass outside interval Shift half mass across interval

Shift some mass across intervaland no further

Shift some mass across interval

Page 52: Joint work with Yuval Peres, Mikkel Thorup, Peter Winkler and Uri Zwick Overhang Bounds Mike Paterson DIMAP & Dept of Computer Science University of Warwick

Simplified setting

“Integral” distributions

Splitting moves

Page 53: Joint work with Yuval Peres, Mikkel Thorup, Peter Winkler and Uri Zwick Overhang Bounds Mike Paterson DIMAP & Dept of Computer Science University of Warwick

0 1 2 3-3 -2 -1

Page 54: Joint work with Yuval Peres, Mikkel Thorup, Peter Winkler and Uri Zwick Overhang Bounds Mike Paterson DIMAP & Dept of Computer Science University of Warwick

Basic challenge

Suppose that we start with a mass of 1 at the origin.

How many splits are needed to get, say, half of the mass to

distance d ?

Reminiscent of a random walk on the line

O(d3) splits are “clearly” sufficient

To prove: Ω (d3) splits are required

Page 55: Joint work with Yuval Peres, Mikkel Thorup, Peter Winkler and Uri Zwick Overhang Bounds Mike Paterson DIMAP & Dept of Computer Science University of Warwick

Effect of a split

Note that such split moves here have associated interval of length 2.

Page 56: Joint work with Yuval Peres, Mikkel Thorup, Peter Winkler and Uri Zwick Overhang Bounds Mike Paterson DIMAP & Dept of Computer Science University of Warwick

Spread vs. second moment argument

Page 57: Joint work with Yuval Peres, Mikkel Thorup, Peter Winkler and Uri Zwick Overhang Bounds Mike Paterson DIMAP & Dept of Computer Science University of Warwick

That’s a start!

Can we extend the proof to the general case, with general distributions and moves?

Can we get improved boundsfor small values of p?

Can moves beyond position d help?

But …

We did not yet use the lossy nature of moves.

Page 58: Joint work with Yuval Peres, Mikkel Thorup, Peter Winkler and Uri Zwick Overhang Bounds Mike Paterson DIMAP & Dept of Computer Science University of Warwick

Spread vs. second moment argument

Page 59: Joint work with Yuval Peres, Mikkel Thorup, Peter Winkler and Uri Zwick Overhang Bounds Mike Paterson DIMAP & Dept of Computer Science University of Warwick

Spread vs. second moment argument

Page 60: Joint work with Yuval Peres, Mikkel Thorup, Peter Winkler and Uri Zwick Overhang Bounds Mike Paterson DIMAP & Dept of Computer Science University of Warwick

Spread vs. second moment inequalities

If 1 is obtained from 0 by an extreme move, then

Plackett (1947):

Page 61: Joint work with Yuval Peres, Mikkel Thorup, Peter Winkler and Uri Zwick Overhang Bounds Mike Paterson DIMAP & Dept of Computer Science University of Warwick

Spread vs. second moment argument(for extreme moves)

Page 62: Joint work with Yuval Peres, Mikkel Thorup, Peter Winkler and Uri Zwick Overhang Bounds Mike Paterson DIMAP & Dept of Computer Science University of Warwick

Splitting

“Basic” splitting move

A single mass is split into arbitrarily many

parts, maintaining the total and center of mass

if 1 is obtained from 0 by a sequence of splitting moves

Def:

Page 63: Joint work with Yuval Peres, Mikkel Thorup, Peter Winkler and Uri Zwick Overhang Bounds Mike Paterson DIMAP & Dept of Computer Science University of Warwick

Splitting and extreme moves

If V is a sequence of moves, we let V* be the corresponding sequence of

extreme moves

Lemma:

Corollary:

Page 64: Joint work with Yuval Peres, Mikkel Thorup, Peter Winkler and Uri Zwick Overhang Bounds Mike Paterson DIMAP & Dept of Computer Science University of Warwick

Spread vs. second moment argument(for general moves)

extreme

Page 65: Joint work with Yuval Peres, Mikkel Thorup, Peter Winkler and Uri Zwick Overhang Bounds Mike Paterson DIMAP & Dept of Computer Science University of Warwick

Notation

Page 66: Joint work with Yuval Peres, Mikkel Thorup, Peter Winkler and Uri Zwick Overhang Bounds Mike Paterson DIMAP & Dept of Computer Science University of Warwick

An extended bound

Page 67: Joint work with Yuval Peres, Mikkel Thorup, Peter Winkler and Uri Zwick Overhang Bounds Mike Paterson DIMAP & Dept of Computer Science University of Warwick

An almost tight bound

Page 68: Joint work with Yuval Peres, Mikkel Thorup, Peter Winkler and Uri Zwick Overhang Bounds Mike Paterson DIMAP & Dept of Computer Science University of Warwick

An almost tight bound - Proof

Page 69: Joint work with Yuval Peres, Mikkel Thorup, Peter Winkler and Uri Zwick Overhang Bounds Mike Paterson DIMAP & Dept of Computer Science University of Warwick

An asymptotically tight bound

lossy moves

Page 70: Joint work with Yuval Peres, Mikkel Thorup, Peter Winkler and Uri Zwick Overhang Bounds Mike Paterson DIMAP & Dept of Computer Science University of Warwick

An asymptotically tight bound - Proof

lossy

Page 71: Joint work with Yuval Peres, Mikkel Thorup, Peter Winkler and Uri Zwick Overhang Bounds Mike Paterson DIMAP & Dept of Computer Science University of Warwick

Our paper was in SODA’08 this week

An early version is at http://arXiv.org/pdf/0707.0093

Page 72: Joint work with Yuval Peres, Mikkel Thorup, Peter Winkler and Uri Zwick Overhang Bounds Mike Paterson DIMAP & Dept of Computer Science University of Warwick

Some open questions

What shape gives optimal overhang?

We only consider frictionless 2D constructions here. This implies no horizontal forces, so, even if blocks are tilted, our results still hold. What happens in the frictionless 3D case?

With friction, everything changes!

Page 73: Joint work with Yuval Peres, Mikkel Thorup, Peter Winkler and Uri Zwick Overhang Bounds Mike Paterson DIMAP & Dept of Computer Science University of Warwick

With friction

With enough friction we can get overhang greater than 1 with only 2 blocks!

With enough friction, all diamonds are balanced, so we get Ω(n1/2) overhang.

Probably we can get Ω(n1/2) overhang with arbitrarily small friction.

With enough friction, there are possibilities to get exponents greater than 1/2.

In 3D, I think that when the coefficient of friction is greater than 1 we can get Ω(n) overhang.

Page 74: Joint work with Yuval Peres, Mikkel Thorup, Peter Winkler and Uri Zwick Overhang Bounds Mike Paterson DIMAP & Dept of Computer Science University of Warwick

The end

Applications!

Page 75: Joint work with Yuval Peres, Mikkel Thorup, Peter Winkler and Uri Zwick Overhang Bounds Mike Paterson DIMAP & Dept of Computer Science University of Warwick

The end