Applied Topology Henry Adams July 18, 2013 Stanford University MathemaAcs Camp Filtration parameter t Datasets have shapes What shape is this? 0.191 Datasets have shapes Example: Diabetes study 145 points in 5‐dimensional space G. M. Reaven and R. G. Miller: The Nature of Chemical Diabetes 19 Fig. 1. Artist's rendition of data as seen in three dimensions. View is approximately along 45~line as seen through Prim 9 program on the computer; coordinate axes are in the 9background An A$empt to Define the Nature of Chemical Diabetes Using a Mul;dimensional Table 1. Classification of the 145 subjects into three groups on the basis of the oral glucose tolerance test Analysis by G. M. Reaven and R. G. Miller, 1979. Metabolic characteristics (mean + SD) Group Number Rel.wt. Glucose area Insulin area SSPG Datasets have shapes Example: Cyclo‐Octane (C8H16) data 1,031,644 points in 72‐dimensional space Non‐Manifold Surface Reconstruc;on from High Dimensional Point Cloud Data by Shawn MarAn and Jean‐Paul Watson, 2010. Datasets have shapes Example: Cyclo‐Octane (C8H16) data 1,031,644 points in 72‐dimensional space Figure 7: Conformation Space of Cyclo-Octane. Here we show how the set of conforma tions of Non‐Manifold Surface Reconstruc;on from High Dimensional Point Cloud Data by cyclo-octane can be represented as a surface in a high dimensional space. On the Shawn MarAn and Jean‐Paul Watson, 2010. left, we show various conformations of cyclo-octane as drawn by PyMol (www.pymol.org) In the center, these conformations are represented by the 3D coordinates of their atoms Datasets have shapes Example: Cyclo‐Octane (C8H16) data 1,031,644 points in 72‐dimensional space Figure 7: Conformation Space of Cyclo-Octane. Here we show how the set of conforma tions of Non‐Manifold Surface Reconstruc;on from High Dimensional Point Cloud Data by cyclo-octane can be represented as a surface in a high dimensional space. On the Shawn MarAn and Jean‐Paul Watson, 2010. left, we show various conformations of cyclo-octane as drawn by PyMol (www.pymol.org) In the center, these conformations are represented by the 3D coordinates of their atoms Datasets have shapes Example: Cyclo‐Octane (C8H16) data 1,031,644 points in 72‐dimensional space Figure 8: Triangulation of Cyclo-Octane Conformation Space. Here we show the triangulation obtained by our surface reconstruction algorithm on the cyclo-octane conformation space. The triangulation was carried out in 24 dimensions, but is shown using the reduced dimensional representation provided by Isomap. Self-intersections are shown in black. Non‐Manifold Surface Reconstruc;on from High Dimensional Point Cloud Data by Shawn MarAn and Jean‐Paul Watson, 2010. Topology studies shapes A donut and coffee mug are “homotopy equivalent.” ~ Topology studies shapes A donut and coffee mug are “homotopy equivalent.” ~ 8):11, 1–18 Singh et al. Topology studies shapes Topological Analysis of Popula;on Ac;vity in Visual Cortex by Singh, Memoli, Ishkhanov, Sapiro, Carlsson, and Ringach, 2008. Topology studies shapes Torus Topology studies shapes Topology studies shapes Klein boale R2 Topology studies shapes Homology ( ) Z/2Z R2 Topology studies shapes Homology ( ) Z/2Z Homology groups H0, H1, H2, H3, … Hk “counts the number of k‐dimensional holes”. Homotopy equivalent shapes have the same homology groups. R2 Topology studies shapes Homology ( ) Z/2Z H0 has rank 1. H1 has rank 1. H2 has rank 0. R2 Topology studies shapes Homology ( ) Z/2Z H0 has rank 1. H1 has rank 3. H2 has rank 0. R2 Topology studies shapes Homology ( ) Z/2Z H0 has rank 1. H1 has rank 0. H2 has rank 1. R2 Topology studies shapes Homology ( ) Z/2Z H0 has rank 3. H1 has rank 4. H2 has rank 1. R2 Topology studies shapes Homology ( ) Z/2Z H0 has rank 1. H1 has rank 2. H2 has rank 1. R2 Topology studies shapes Homology ( ) Z/2Z H0 has rank 1. H1 has rank 2. H2 has rank 1. R2 Topology studies shapes Homology ( ) Z/2Z 0‐simplex 1‐simplex 2‐simplex 3‐simplex R2 Topology studies shapes Homology ( ) Z/2Z 0‐simplex 1‐simplex 2‐simplex Simplicial complexes 3‐simplex R2 Topology studies shapes Homology ( ) Z/2Z 0‐simplex 1‐simplex 2‐simplex 3‐simplex R2 Topology studies shapes Homology ( ) Z/2Z 0‐simplex 1‐simplex 2‐simplex 3‐simplex R2 Topology studies shapes Homology ( ) Z/2Z 0‐simplices 1‐simplices 2‐simplices R2 Topology studies shapes Homology ( ) Z/2Z 0‐simplices 1‐simplices 2‐simplices R2 Topology studies shapes Homology ( ) Z/2Z 0‐cycle 1‐cycle A cycle has no boundary. 2‐cycle R2 Topology studies shapes Homology ( ) Z/2Z 0‐cycle 1‐cycle A cycle has no boundary. 2‐cycle + Bp with c ∈ Z . More formally, this collection is ca p R2 ycles in the same coset are said to be homologous, whic Topology studies shapes see Figure IV.5. We may take c as the representative Homology ( ) Z/2Z A torus with three closed curves, each a 1-cycle. Only one Two cycles are equivalent if they differ by a boundary. mologous to the sum of the other two. The sum of the t Hk measures equivalence classes of k‐cycles. 1-boundary, namely of the pair of pants between them. R2 Topology studies shapes Homology ( ) Z/2Z H0 has rank 1. H1 has rank 1. H2 has rank 0. Two cycles are equivalent if they differ by a boundary. Hk measures equivalence classes of k‐cycles. R2 Topology studies shapes Homology ( ) Z/2Z Homology groups H0, H1, H2, H3, … Hk “counts the number of k‐dimensional holes”. Homotopy equivalent shapes have the same homology groups. “Topology! The stratosphere of human thought! In the twenty‐fourth century it might possibly be of use to someone …” ‐Aleksandr Solzhenitsyn, The First Circle Appearance draw balls Topology applied to data analysis draw complex Filtration parameter What shape is this? t 0.191 Appearance draw balls Topology applied to data analysis draw complex Filtration parameter What shape is this? t 0.191 Appearance draw balls Topology applied to data analysis draw complex Filtration parameter What shape is this? Cech complex t 0.191 Appearance draw balls Topology applied to data analysis draw complex Filtration parameter What shape is this? Cech complex t 0.191 Appearance draw balls Topology applied to data analysis draw complex Filtration parameter What shape is this? Cech complex t 0.191 Topology applied to data analysis What shape is this? Cech complex Topology applied to data analysis What shape is this? Cech complex Topology applied to data analysis Significant features persist. Image patch example Study 3x3 high‐contrast patches from images Points in 9‐dimensional space On the Local Behavior of Spaces of Natural Images by Gunnar Carlsson, Tigran Ishkhanov, Vin de Silva, and Afra Zomorodian, 2008. Image patch example 1st densest group of patches Image patch example 1st densest group of patches "#% "#& "#' "#( " !"#( !"#' !"#& !"#% !! !"#$ " "#$ ! InterpretaAon: nature prefers linearity Image patch example 2nd densest group of patches Image patch example 2nd densest group of patches "#% "#& "#' "#( " !"#( !"#' !"#& !"#% !! !"#$ " "#$ ! InterpretaAon: nature prefers horizontal and verAcal direcAons Image patch example 3rd densest group of patches 7 Image patch example 3rd densest group of patches Image patch example 3rd densest group of patches Image patch example 3rd densest group of patches Image patch example 3rd densest group of patches Image patch example 3rd densest group of patches Image patch example 3rd densest group of patches Image patch example 3rd densest group of patches InterpretaAon: nature prefers linear and quadraAc patches at all angles Datasets have shapes Example: Cyclo‐Octane (C8H16) data 1,031,644 points in 72‐dimensional space Figure 8: Triangulation of Cyclo-Octane Conformation Space. Here we show the triangulation obtained by our surface reconstruction algorithm on the cyclo-octane conformation space. The triangulation was carried out in 24 dimensions, but is shown using the reduced dimensional representation provided by Isomap. Self-intersections are shown in black. Non‐Manifold Surface Reconstruc;on from High Dimensional Point Cloud Data by Shawn MarAn and Jean‐Paul Watson, 2010. 4 a.nb In[422]:= Evasion paths in mobile sensor networks 12 Rips!data1, t, 0, 1" Rips simplicial complex Appearance draw complex draw balls Filtration parameter In[423]:= In[429]:= Rips!data3n, t, 0, 1" In[430]:= a.nb Rips!data4n, t, 0, 1" 5 In[431]:= a.nb Rips!data5n, t, 0, 1" 11 In[427]:= 10 Rips!data6, t, 0, 1" a.nb In[428]:= a.nb Rips simplicial complex Rips simplicial complex Rips simplicial complex Rips simplicial complex Rips simplicial complex Appearance Appearance Appearance Appearance Appearance Appearance draw complex draw complex draw complex draw complex draw complex draw complex draw balls draw balls draw balls draw balls draw balls draw balls Filtration parameter t0.358 Out[423]= Out[429]= Printed by Mathematica for Students Filtration parameter t0.358 Out[430]= Printed by Mathematica for Students Filtration parameter t0.358 Out[431]= Printed by Mathematica for Students Filtration parameter t0.358 Printed by Mathematica for Students a.nb Filtration parameter t0.358 Out[427]= 13 Rips data7, t, 0, 1 Rips simplicial complex Filtration parameter t Out[422]= Rips data2, t, 0, 1 a.nb t0.358 0.358 Out[428]= Printed by Mathematica for Students Printed by Mathematica for Students Printed by Mathematica for Students 9 Coverage problem • Sensors in a domain D ⊂ Rn . Coordinate-free Coverage in Sensor Networks with Controlled Boundaries via Homology by V. de Silva and R. Ghrist Coverage problem • Sensors in a domain D ⊂ Rn . • Measure only local connectivity. Coordinate-free Coverage in Sensor Networks with Controlled Boundaries via Homology by V. de Silva and R. Ghrist Coverage problem • Sensors in a domain D ⊂ Rn . • Measure only local connectivity. • Čech complex. Coordinate-free Coverage in Sensor Networks with Controlled Boundaries via Homology by V. de Silva and R. Ghrist Coverage problem • Sensors in a domain D ⊂ Rn . • Measure only local connectivity. • Čech complex. Coordinate-free Coverage in Sensor Networks with Controlled Boundaries via Homology by V. de Silva and R. Ghrist Evasion problem • Sensors move in a domain D ⊂ Rn over time interval I . Coordinate-free Coverage in Sensor Networks with Controlled Boundaries via Homology by V. de Silva and R. Ghrist Evasion problem • Sensors move in a domain D ⊂ Rn over time interval I . Coordinate-free Coverage in Sensor Networks with Controlled Boundaries via Homology by V. de Silva and R. Ghrist Evasion problem • Sensors move in a domain D ⊂ Rn over time interval I . • Measure only local connectivity. Coordinate-free Coverage in Sensor Networks with Controlled Boundaries via Homology by V. de Silva and R. Ghrist Evasion problem • Sensors move in a domain D ⊂ Rn over time interval I . • Measure only local connectivity. • Čech complex. Coordinate-free Coverage in Sensor Networks with Controlled Boundaries via Homology by V. de Silva and R. Ghrist Evasion problem • Sensors move in a domain D ⊂ Rn over time interval I . • Measure only local connectivity. • Čech complex. Coordinate-free Coverage in Sensor Networks with Controlled Boundaries via Homology by V. de Silva and R. Ghrist Evasion problem • Let X ⊂ D × I be the covered region. D I Coordinate-free Coverage in Sensor Networks with Controlled Boundaries via Homology by V. de Silva and R. Ghrist Evasion problem • Let X ⊂ D × I be the covered region. • Using coordinate-free input, can we determine if an evasion path exists? D I Coordinate-free Coverage in Sensor Networks with Controlled Boundaries via Homology by V. de Silva and R. Ghrist Evasion problem • Let X ⊂ D × I be the covered region. • Using coordinate-free input, can we determine if an evasion path exists? D I Coordinate-free Coverage in Sensor Networks with Controlled Boundaries via Homology by V. de Silva and R. Ghrist Evasion problem 4 • Let X ⊂ D × I be the covered region. • Using coordinate-free input, can we determine if an evasion path exists? a.nb In[422]:= 6 Rips data1, t, 0, 1 In[423]:= In[424]:= a.nb 8 5 a.nb Rips!data3y, t, 0, 1" In[425]:= Rips!data4y, t, 0, 1" In[426]:= a.nb 107 Rips!data5y, t, 0, 1" In[427]:= Rips!data6, t, 0, 1" a.nb In[428]:= Rips simplicial complex Rips simplicial complex Rips simplicial complex Rips simplicial complex Rips simplicial complex Rips simplicial complex Appearance Appearance Appearance Appearance Appearance Appearance Appearance draw complex draw complex draw complex draw complex draw complex draw complex draw complex draw balls draw balls draw balls draw balls draw balls draw balls draw balls Filtration parameter t Filtration parameter t0.358 Out[423]= Filtration parameter t0.358 Out[424]= Filtration parameter t0.358 Out[425]= Filtration parameter t0.358 Out[426]= Filtration parameter t0.358 Out[427]= a.nb Rips data7, t, 0, 1 Rips simplicial complex Filtration parameter Out[422]= Rips data2, t, 0, 1 a.nb t0.358 0.358 Out[428]= Coordinate-free Coverage in Sensor Networks with Controlled Boundaries via Homology by V. de Silva and R. Ghrist Printed by Mathematica for Students Printed by Mathematica for Students Printed by Mathematica for Students Printed by Mathematica for Students Printed by Mathematica for Students Printed by Mathematica for Students Printed by Mathematica for Students 9 Evasion problem • Stacked complex SC ! X encodes all Čech complexes. Coordinate-free Coverage in Sensor Networks with Controlled Boundaries via Homology by V. de Silva and R. Ghrist Evasion problem • Stacked complex SC ! X encodes all Čech complexes. Coordinate-free Coverage in Sensor Networks with Controlled Boundaries via Homology by V. de Silva and R. Ghrist Evasion problem • Stacked complex SC ! X encodes all Čech complexes. Coordinate-free Coverage in Sensor Networks with Controlled Boundaries via Homology by V. de Silva and R. Ghrist Evasion problem • Theorem 7 (reformulated) If there is an α ∈ Hn (SC, ∂D × I) with 0 != δα ∈ Hn−1 (∂D × I) , then no evasion path exists. D α I Coordinate-free Coverage in Sensor Networks with Controlled Boundaries via Homology by V. de Silva and R. Ghrist Evasion problem • Theorem 7 (reformulated) If there is an α ∈ Hn (SC, ∂D × I) with 0 != δα ∈ Hn−1 (∂D × I) , then no evasion path exists. D I Coordinate-free Coverage in Sensor Networks with Controlled Boundaries via Homology by V. de Silva and R. Ghrist Evasion problem • Theorem 7 (reformulated) If there is an α ∈ Hn (SC, ∂D × I) with 0 != δα ∈ Hn−1 (∂D × I) , then no evasion path exists. D I Coordinate-free Coverage in Sensor Networks with Controlled Boundaries via Homology by V. de Silva and R. Ghrist Evasion problem • Theorem 7 (reformulated) If there is an α ∈ Hn (SC, ∂D × I) with 0 != δα ∈ Hn−1 (∂D × I) , then no evasion path exists. • Coordinate-free. • Not sharp. Can it be sharpened? D I Coordinate-free Coverage in Sensor Networks with Controlled Boundaries via Homology by V. de Silva and R. Ghrist Evasion problem • Theorem 7 (reformulated) If there is an α ∈ Hn (SC, ∂D × I) with 0 != δα ∈ Hn−1 (∂D × I) , then no evasion path exists. • Coordinate-free. • Not sharp. Can it be sharpened? D I Coordinate-free Coverage in Sensor Networks with Controlled Boundaries via Homology by V. de Silva and R. Ghrist Zigzag persistence • Form ZSC and take Hn−1 (ZSC) ∼ = Hn−1 (ZX). Zigzag Persistence by G. Carlsson and V. de Silva Zigzag persistence !→ !→ !→ Zigzag Persistence by G. Carlsson and V. de Silva ←! ←! ←! ←! !→ • Form ZSC and take Hn−1 (ZSC) ∼ = Hn−1 (ZX). Zigzag persistence • Form ZSC and take Hn−1 (ZSC) ∼ = Hn−1 (ZX). Zigzag Persistence by G. Carlsson and V. de Silva Zigzag persistence • Form ZSC and take Hn−1 (ZSC) ∼ = Hn−1 (ZX). • Hypothesis: there is an evasion path ⇔ there is a long bar. Zigzag Persistence by G. Carlsson and V. de Silva Zigzag persistence • Form ZSC and take Hn−1 (ZSC) ∼ = Hn−1 (ZX). • Hypothesis: there is an evasion path ⇔ there is a long bar. • ⇒ is true, but ⇐ is false. Zigzag Persistence by G. Carlsson and V. de Silva Dependence on embedding X !→ D × I • SC alone does not determine if an evasion path exists. • Two networks with the same SC . Top contains evasion path while bottom does not. Dependence on embedding X !→ D × I 4 a.nb In[422]:= • SC alone does not determine if an evasion path exists. • Two networks with the same SC . Top contains evasion path while bottom does not. 6 Rips data1, t, 0, 1 In[423]:= In[424]:= a.nb 8 5 a.nb Rips!data3y, t, 0, 1" In[425]:= Rips!data4y, t, 0, 1" In[426]:= a.nb 107 Rips!data5y, t, 0, 1" In[427]:= Rips!data6, t, 0, 1" a.nb In[428]:= Rips simplicial complex Rips simplicial complex Rips simplicial complex Rips simplicial complex Rips simplicial complex Rips simplicial complex Appearance Appearance Appearance Appearance Appearance Appearance Appearance draw complex draw complex draw complex draw complex draw complex draw complex draw complex draw balls draw balls draw balls draw balls draw balls draw balls draw balls Filtration parameter t Filtration parameter t0.358 Out[423]= Out[424]= Printed by Mathematica for Students Filtration parameter t0.358 Out[425]= Printed by Mathematica for Students Filtration parameter t0.358 Out[426]= Printed by Mathematica for Students Filtration parameter t0.358 Printed by Mathematica for Students Filtration parameter t0.358 Out[427]= a.nb Rips data7, t, 0, 1 Rips simplicial complex Filtration parameter Out[422]= Rips data2, t, 0, 1 a.nb t0.358 0.358 Out[428]= Printed by Mathematica for Students Printed by Mathematica for Students Printed by Mathematica for Students 9 Dependence on embedding X !→ D × I 4 a.nb In[422]:= • SC alone does not determine if an evasion path exists. • Two networks with the same SC . Top contains evasion path while bottom does not. 6 Rips data1, t, 0, 1 In[423]:= Rips!data3y, t, 0, 1" In[425]:= Rips!data4y, t, 0, 1" In[426]:= a.nb 107 Rips!data5y, t, 0, 1" In[427]:= Rips!data6, t, 0, 1" a.nb In[428]:= Rips simplicial complex Rips simplicial complex Rips simplicial complex Rips simplicial complex Rips simplicial complex Rips simplicial complex Appearance Appearance Appearance Appearance Appearance Appearance draw complex draw complex draw complex draw complex draw complex draw complex draw complex draw balls draw balls draw balls draw balls draw balls draw balls draw balls Filtration parameter 4Out[422]= a.nb In[423]:= Rips data2, t, 0, 1 Filtration parameter t0.358 6Out[424]= a.nb Out[423]= Rips data1, t, 0, 1 Filtration parameter t0.358 In[424]:= In[425]:= Filtration parameter t0.358 a.nb 8Out[426]= 5 a.nb Out[425]= Rips!data3y, t, 0, 1" Filtration parameter t0.358 Rips!data4y, t, 0, 1" In[426]:= Rips!data5y, t, 0, 1" In[427]:= t0.358 Rips!data6, t, 0, 1" In[428]:= Rips data7, t, 0, 1 Rips simplicial complex Rips simplicial complex Rips simplicial complex Rips simplicial complex Rips simplicial complex Rips simplicial complex Appearance Appearance Appearance Appearance Appearance Appearance Appearance draw complex draw complex draw complex draw complex draw complex draw complex draw complex draw balls draw balls draw balls draw balls draw balls draw balls Filtration parameter t Filtration parameter t0.358 Out[423]= t0.358 Out[424]= Printed by Mathematica for Students Filtration parameter t0.358 Out[425]= Printed by Mathematica for Students Filtration parameter Out[426]= Printed by Mathematica for Students Filtration parameter t0.358 Out[427]= Printed by Mathematica for Students draw balls Filtration parameter t0.358 t0.358 0.358 Out[428]= Printed by Mathematica for Students a.nb 9 0.358 a.nb 10 7 a.nb Out[428]= Rips simplicial complex Filtration parameter 9 Filtration parameter t0.358 Out[427]= a.nb Rips data7, t, 0, 1 Appearance t Out[422]= In[424]:= a.nb 8 5 a.nb Rips simplicial complex Filtration parameter In[422]:= Rips data2, t, 0, 1 a.nb Printed by Mathematica for Students Printed by Mathematica for Students Dependence on embedding X !→ D × I 4 a.nb In[422]:= • SC alone does not determine if an evasion path exists. • Two networks with the same SC . Top contains evasion path while bottom does not. 6 Rips data1, t, 0, 1 In[423]:= Rips!data3y, t, 0, 1" In[425]:= Rips!data4y, t, 0, 1" In[426]:= a.nb 107 Rips!data5y, t, 0, 1" In[427]:= Rips!data6, t, 0, 1" a.nb In[428]:= Rips simplicial complex Rips simplicial complex Rips simplicial complex Rips simplicial complex Rips simplicial complex Rips simplicial complex Appearance Appearance Appearance Appearance Appearance Appearance draw complex draw complex draw complex draw complex draw complex draw complex draw complex draw balls draw balls draw balls draw balls draw balls draw balls draw balls Filtration parameter 4Out[422]= a.nb Out[423]= Rips!data1, t, 0, 1" Filtration parameter t0.358 In[423]:= In[429]:= Rips!data3n, t, 0, 1" Filtration parameter t0.358 12 a.nb Out[425]= Out[424]= Rips data2, t, 0, 1 Filtration parameter t0.358 In[430]:= Filtration parameter t0.358 a.nb Out[426]= 5 Rips!data4n, t, 0, 1" In[431]:= Rips!data5n, t, 0, 1" 11 Out[427]= In[427]:= t0.358 10 a.nb Out[428]= Rips!data6, t, 0, 1" In[428]:= a.nb Rips simplicial complex Rips simplicial complex Rips simplicial complex Rips simplicial complex Rips simplicial complex Rips simplicial complex Appearance Appearance Appearance Appearance Appearance Appearance draw complex draw complex draw complex draw complex draw complex draw complex draw complex draw balls draw balls draw balls draw balls draw balls draw balls t Filtration parameter t0.358 Out[423]= t0.358 Out[429]= Printed by Mathematica for Students Filtration parameter t0.358 Out[430]= Printed by Mathematica for Students Filtration parameter Out[431]= Printed by Mathematica for Students Filtration parameter t0.358 Out[427]= Printed by Mathematica for Students draw balls Filtration parameter t0.358 13 Rips data7, t, 0, 1 Appearance Filtration parameter t0.358 0.358 Out[428]= Printed by Mathematica for Students a.nb 9 0.358 Rips simplicial complex Filtration parameter 9 Filtration parameter t0.358 a.nb a.nb Rips data7, t, 0, 1 Appearance t Out[422]= In[424]:= a.nb 8 5 a.nb Rips simplicial complex Filtration parameter In[422]:= Rips data2, t, 0, 1 a.nb Printed by Mathematica for Students Printed by Mathematica for Students Dependence on embedding X !→ D × I 4 a.nb In[422]:= 6 Rips data1, t, 0, 1 In[423]:= Rips!data3y, t, 0, 1" In[425]:= Rips!data4y, t, 0, 1" In[426]:= a.nb 107 Rips!data5y, t, 0, 1" In[427]:= Rips!data6, t, 0, 1" a.nb In[428]:= Rips simplicial complex Rips simplicial complex Rips simplicial complex Rips simplicial complex Rips simplicial complex Rips simplicial complex Appearance Appearance Appearance Appearance Appearance Appearance draw complex draw complex draw complex draw complex draw complex draw complex draw complex draw balls draw balls draw balls draw balls draw balls draw balls draw balls Filtration parameter 4Out[422]= a.nb Out[423]= Rips!data1, t, 0, 1" Filtration parameter t0.358 In[423]:= In[429]:= Rips!data3n, t, 0, 1" Filtration parameter t0.358 12 a.nb Out[425]= Out[424]= Rips data2, t, 0, 1 Filtration parameter t0.358 In[430]:= Filtration parameter t0.358 a.nb Out[426]= 5 Rips!data4n, t, 0, 1" In[431]:= Rips!data5n, t, 0, 1" 11 Out[427]= In[427]:= t0.358 10 a.nb Out[428]= Rips!data6, t, 0, 1" In[428]:= a.nb Rips simplicial complex Rips simplicial complex Rips simplicial complex Rips simplicial complex Rips simplicial complex Rips simplicial complex Appearance Appearance Appearance Appearance Appearance Appearance draw complex draw complex draw complex draw complex draw complex draw complex draw complex draw balls draw balls draw balls draw balls draw balls draw balls t Filtration parameter t0.358 Out[423]= t0.358 Out[429]= Printed by Mathematica for Students Filtration parameter t0.358 Out[430]= Printed by Mathematica for Students Filtration parameter Out[431]= Printed by Mathematica for Students Filtration parameter t0.358 Out[427]= Printed by Mathematica for Students draw balls Filtration parameter t0.358 13 Rips data7, t, 0, 1 Appearance Filtration parameter t0.358 0.358 Out[428]= Printed by Mathematica for Students a.nb 9 0.358 Rips simplicial complex Filtration parameter 9 Filtration parameter t0.358 a.nb a.nb Rips data7, t, 0, 1 Appearance t Out[422]= In[424]:= a.nb 8 5 a.nb Rips simplicial complex Filtration parameter In[422]:= Rips data2, t, 0, 1 a.nb Printed by Mathematica for Students Printed by Mathematica for Students Dependence on embedding X !→ D × I • The two covered regions are fibrewise homotopy equivalent but their complements are not. Zigzag persistence Fat graphs • What minimal sensing capabilities might we add? 2 • Let D ⊂ R . A fat graph structure specifies the cyclic ordering of edges adjacent to each vertex. • Equivalent to a set of boundary cycles. Fat graphs A Delaunay triangulation, with four • The alpha complex and fat graphofstructure ofcentered a connected intersection the !-ball at p and the Voron sensor network determine if anofevasion path exists. the union !-balls centered at points in P. The alph the set {B̌! (p) | p ∈ P}. The underlying space |α! (P)| i immediately implies that the alpha shape is homotop Edelsbrunner [26] for a self-contained proof. If the point set P is in general position, the alp intersection of the Delaunay triangulation of P and th points in P define a simplex in the alpha complex if an at most ! that contains no other point in P. Aleksandrov-Čech complexes and unions of balls for two differe 15.1.2 Vietoris-Rips Complexes: Flags and Shadow The proximity graph N! (P) is the geometric graph who all pairs of points at distance at most 2!; in other words, complex. The Vietoris-Rips complex VR! (P) is the fla An alpha complex and a decomposed union of balls. graph N! (P). A set of k + 1 points in P defines a k-simpl Out[24]= Fat graphs 4 FatGraphHard_edit.nb FatGraphHard_edit.nb 3 • The alpha complex and fat graph structure of a connected sensor network determine if an evasion path exists. Cech simplicial complex Cech simplicial complex Appearance Appearance draw one simplices draw one simplices draw TWEAKED one simplices draw TWEAKED one simplices draw Cech complex draw Cech complex Filtration parameter Filtration parameter t t 0.249 Out[24]= 0.249 Fat graphs • The alpha complex and fat graph structure of a connected sensor network determine if an evasion path exists. Fat graphs • The alpha complex and fat graph structure of a connected sensor network determine if an evasion path exists. Fat graphs • The alpha complex and fat graph structure of a connected sensor network determine if an evasion path exists. Fat graphs • The alpha complex and fat graph structure of a connected sensor network determine if an evasion path exists. Fat graphs • The alpha complex and fat graph structure of a connected sensor network determine if an evasion path exists. Fat graphs • The alpha complex and fat graph structure of a connected sensor network determine if an evasion path exists. Fat graphs • The alpha complex and fat graph structure of a connected sensor network determine if an evasion path exists. Fat graphs • The alpha complex and fat graph structure of a connected sensor network determine if an evasion path exists. Fat graphs • The alpha complex and fat graph structure of a connected sensor network determine if an evasion path exists. Rips simplicial complex Rips simplicial complex Rips simplicial complex Rips simplicial complex Rips simplicial complex Rips simplicial complex Rips simplicial complex Appearance Appearance Appearance Appearance Appearance Appearance Appearance draw complex draw complex draw complex draw complex draw complex draw complex draw complex draw balls draw balls draw balls draw balls draw balls draw balls Filtration parameter Filtration parameter t 0.364 Filtration parameter t 0.364 t Fat graphs Filtration parameter 0.364 t 0.364 draw balls Filtration parameter Filtration parameter t Filtration parameter t 0.364 0.364 t • The alpha complex and fat graph structure of a connected sensor network determine if an evasion path exists. Printed by Mathematica for Students 2 Printed by Mathematica for Students 4 bNo.nb bYes.nb bNo.nb Printed by Mathematica for Students 6 3 Printed by Mathematica for Students bYes.nb bYes.nb Printed by Mathematica for Students 5 8 Printed by Mathematica for Students bNo.nb bNo.nb Rips simplicial complex Rips simplicial complex Rips simplicial complex Rips simplicial complex Rips simplicial complex Rips simplicial complex Rips simplicial complex Appearance Appearance Appearance Appearance Appearance Appearance Appearance draw complex draw complex draw complex draw complex draw complex draw complex draw complex draw balls Filtration parameter t 0.364 draw balls Filtration parameter Filtration parameter t 4 0.364 bNo.nb 0.364 6 0.364 bNo.nb Filtration parameter t Printed by Mathematica for Students bNo.nb draw balls Filtration parameter t Printed by Mathematica for Students 3 draw balls Filtration parameter t Printed by Mathematica for Students bNo.nb draw balls Filtration parameter t 0.364 Printed by Mathematica for Students bNo.nb 0.364 t Printed by Mathematica for Students 5 8 bNo.nb bNo.nb Rips simplicial complex Rips simplicial complex Rips simplicial complex Rips simplicial complex Rips simplicial complex Rips simplicial complex Appearance Appearance Appearance Appearance Appearance Appearance Appearance draw complex draw complex draw complex draw complex draw complex draw complex draw complex draw balls draw balls draw balls draw balls draw balls draw balls t Filtration parameter 0.364 Printed by Mathematica for Students t Filtration parameter 0.364 Printed by Mathematica for Students t Filtration parameter 0.364 Printed by Mathematica for Students t Filtration parameter 0.364 Printed by Mathematica for Students t Filtration parameter 0.364 Printed by Mathematica for Students t 0.364 Printed by Mathematica for Students Rips simplicial complex Filtration parameter 7 Pair A: Top row contains evasion path; bottom does not. draw balls draw balls Printed by Mathematica for Students 2 Printed by Mathematica for Students 0.364 7 draw balls Filtration parameter 0.364 t Printed by Mathematica for Students 0.364 Printed by Mathematica for Students Pair B: Top row contains evasion path; bottom does not. hat minimal sensor capabilites might one add to distinguish these examples? Each covered region Out[24]= Fat graphs 4 FatGraphHard_edit.nb FatGraphHard_edit.nb 3 • The alpha complex and fat graph structure of a connected sensor network determine if an evasion path exists. • Are the Čech complex and fat graph structure sufficient? Cech simplicial complex Cech simplicial complex Appearance Appearance draw one simplices draw one simplices draw TWEAKED one simplices draw TWEAKED one simplices draw Cech complex draw Cech complex Filtration parameter Filtration parameter t t 0.249 Out[24]= 0.249 Fat graphs • The alpha complex and fat graph structure of a connected sensor network determine if an evasion path exists. • Are the Čech complex and fat graph structure sufficient? $ & " # # & % % ! $ " ! 4 a.nb In[422]:= 12 Rips!data1, t, 0, 1" In[423]:= In[429]:= Rips!data3n, t, 0, 1" In[430]:= a.nb Rips!data4n, t, 0, 1" 5 In[431]:= a.nb Rips!data5n, t, 0, 1" 11 In[427]:= 10 Rips!data6, t, 0, 1" a.nb In[428]:= a.nb a.nb • Final thought: ideas from pure mathemaAcs are oien helpful in tackling problems arising from more applied sejngs. Rips simplicial complex Rips simplicial complex Rips simplicial complex Rips simplicial complex Rips simplicial complex Rips simplicial complex Appearance Appearance Appearance Appearance Appearance Appearance Appearance draw complex draw complex draw complex draw complex draw complex draw complex draw balls draw balls draw balls draw balls draw balls draw balls draw balls 13 Rips data7, t, 0, 1 Rips simplicial complex draw complex Filtration parameter Filtration parameter t Out[422]= Rips data2, t, 0, 1 a.nb Filtration parameter t0.358 Out[423]= Filtration parameter t0.358 Filtration parameter t0.358 Out[429]= Out[430]= Filtration parameter t0.358 Out[431]= Filtration parameter t0.358 Out[427]= t0.358 0.358 Out[428]= Thank you! Printed by Mathematica for Students Printed by Mathematica for Students Printed by Mathematica for Students Printed by Mathematica for Students Printed by Mathematica for Students Printed by Mathematica for Students Printed by Mathematica for Students 9