Lecture 12: Feb 16, 2017

Transcription

1 CS 6170 Computational Topology: Topological Data Analysis Spring 2017 Lecture 12: Feb 16, 2017 Lecturer: Prof Bei Wang University of Utah School of Computing Scribe: Waiming Tai This lecture s notes illustrate the concept of computing homology 121 Review of definitions For any simplicial complex K, we have the following definitions Definition 121 The p-th cycle group Z p (K) is a set of p-th chain C p (K) with empty boundary That is, Z p (K) = {c c = 0, c Z p (K)} Definition 122 The p-th boundary group B p (K) is a set of p-th chain C p (K) that is the boundary of a (p + 1)-th chain That is, B p (K) = {c c = d for some d C p+1 (K)} Definition 123 The p-th homology group H p (K) is the p-th cycle group Z p (K) modulo the p-th boundary group B p That is, H p (K) = Z p (K)/B p (K) From now on, we simplify the notation Z p (K) = Z p, B p (K) = B p and H p (K) = H p when K is apparent Roughly speaking, H p is the group of cycles that don t bound Here is an example Figure 121: The first example Let c = c is a cycle which means c Z 1 However, there is not a d C 2 such that c = d and so c / B 1 Therefore, c is a non-identity element of H

2 Lecture 12: Feb 16, Let c = c Z 1 Also, c = d where d = 123 and so c B 1 That means c is an identity in H 1 Let c = We can express c as ( ) + ( ) = c + c It means that c c in H 1 Here is another example Figure 122: The second example Consider Is this cycle an identity in H 1? The answer is yes We can express it as ( ) + ( ) + ( ) It is easy to see that and are in B 1 but is not Definition 124 A generating set of a group G is a subset of G such that every element in G can be expressed as the combination (under group operation) of finitely many elements of the subset and their inverses Definition 125 Rank of a group G rank(g) is the smallest cardinality of a generating set of G That is, rank(g) = min S G S where minimum is over all generating set of G Definition 126 The p-th Betti number β p is the rank of H p That is, β p = rank(h p )

3 Lecture 12: Feb 16, Figure 123: Generating set example In the above example, rank(h 1 ) = 2 not 3 Consider c 1 = c 2 = c 3 = It is easy to check that the smallest set of H 1 is {c 1, c 2 } or {c 2, c 3 } or {c 1, c 3 } This example also shows that the smallest generating set may not be unique Recall that all p-th chain C p are connected by boundary operator If 123 C 2, then C 2 C1 C0 (123) = C 1 (12) = C 0 More generally, p+1 p C p+1 Cp Cp 1

4 Lecture 12: Feb 16, Figure 124: Illustration of boundary map 122 Reduced homology Consider the augmentation map E : C 0 Z 2 defined by E(u) = 1 for every vertex u C E 0 C1 C0 Z2 = C 1 0 Definition 127 The p-th reduced homology group H p is defined as following H p = ker p \ im p+1 = H p In particular, H 0 = ker E\ im 1 Definition 128 The p-th reduced Betti number β p is the rank of H p That is, β p = rank( H p ) If K is not empty, then { βp = β p, for p 1 β 0 = β 0 1 If K =, then β 1 = Algorithm This is the algorithm for computing β p Input: p-th boundary matrix p for all p where the column represent p-simplices, η p and the row represent (p 1)-simplices, η p 1 Use row and column operation to reduce p to Smith normal form (SNF) N p return n 0 n 1 where n 0 is number of zero column in N p and n 1 is number of non-zero row in N p+1

5 Lecture 12: Feb 16, Recall that a matrix is SNF if all non-diagonal element are zero all non-zero row are above all zero row Indeed, we can prove that n 0 = rank(z p ) and n 1 = rank(b p ) and therefore the output is exactly β p Recall that column and row operation consist of the following Column operation: exchange column k with column l add column k to column l col k col k + col l = col k col l 1 (row k) 1 (col l) 1 Row operation: exchange row k with row l add row l to row k row k + row l row l = 1 (col k) 1 (row l) 1 row k row l Therefore, N p = U p 1 p V p where U p 1 represent the row operation and V p represent the column operation Here is the example The following K is called triangulated 3-ball which consists of all possible combination That is, K = {a, b, c, d, ab, ac, ad, bc, bd, cd, abc, abd, acd, bcd, abcd}

6 Lecture 12: Feb 16, Figure 125: Triangulated 3-ball It is easy to check β 0 = β 0 1 = 0, β 1 = β 1 = 0, β 2 = β 2 = 0 Now, we can compute this by the above algorithm 0 : Adding column 1 to column 2, 3 and 4: N 0 : Therefore, rank(z 0 ) = 3 and rank(b 1 ) = 1 1 : Adding row 1 to row 2: Adding column 1 to column 2 and 3: a b c d a b c d a b c d a b c a b c b c

7 Lecture 12: Feb 16, Adding row 2 to row 3: Adding column 2 to column 3, 4 and 5: Adding row 3 to row 4: N 1 : b c b c b c d b c d Therefore, rank(z 1 ) = 3 and rank(b 0 ) = 3 Also, β 0 = 0 2 : Adding row 1 to row 2 and 4: Adding column 1 to column 2: ab ac ad bc bd ab ac ad bc bd ac ad bc bd

8 Lecture 12: Feb 16, Adding row 2 to row 3, 4, 5 and 6: Adding column 2 to column 3: Exchanging row 3 with row 4: Adding row 3 to row 5 and 6: Adding column 3 to column 4: N 2 : ac bc bd bc bd bc bd bc bd cd bc bd cd bc bd cd Therefore, rank(z 2 ) = 1 and rank(b 1 ) = 3 Also, β 1 = 0

9 Lecture 12: Feb 16, : Adding row 1 to row 2, 3 and 4: N 3 : abcd abc 1 abd 1 acd 1 bcd 1 abcd abc 1 abd 0 acd 0 bcd 0 abcd abc 1 abd 0 acd 0 bcd 0 Therefore, rank(z 3 ) = 0 and rank(b 2 ) = 1 Also, β 2 = 0 Here is another example This example is same as the previous one except that the center is hollow That is, K = {a, b, c, d, ab, ac, ad, bc, bd, cd, abc, abd, acd, bcd} Figure 126: Hollow triangulated 3-ball In this case, 3 doesn t exist Therefore, β 2 = 1 0 = 1