, Given an n × m binary matrix M , it is possible to approximate it by a totally balanced one with the following algorithm: 1. Reorder the lines and the columns of M such that M is doubly lexically ordered, 2. Check and approximate if necessary the doubly lexically ordered matrix M

, Each step can be done in O(nm) operations: Step 1 is an algorithm of Spinrad, 1987.

, our aim is to build a dissimilarity d such that: d(x, y) = min{? j | M x,j = M y,j = 1}

, We suppose without loss of generality that the last column of M is full of 1's. Computing this dissimilarity can be done in O(n 3 ) operations using Algorithm 5, as shown in Proposition 15, In addition, since M is ? -free, d is totally balanced

, Given an n × m ? -free binary matrix M with last column full of 1's and m real positive numbers, Algorithm 5 computes a dissimilarity d such that d(x, y) = min{? j | M x,j = M y,j =

, Notice that for any i, j, N i [j] = k?i M i,j and that there is no repetition in P i . Thus, |P i | ? n for any i ? n

. Since-m-is-?--free, M i ,j = 1}, the lists N i and the lists P i are such that: ? If M i,j = M i,j = 1, N i [j] < N i [j ] ?? {i} ? C ij C ij (for all i, {C i,j : 1 ? j ? m, M i,j = 1} is a chain); this allows to check wether C ij ? C ij by comparing N i

?. If-j-&lt;-j-and-m-i,j-=-m-i,j-=-1,-then-c-ij-?-c-ij, So after Line 14, P i contains all indices of the interesting columns (and only them): -Indices not in P i are useless: if, for j / ? P i , M i,j = M i ,j = 1, then there exists j ? P i with M i,j = M i ,j = 1 and ? j < ? j . -All indices in P i are useful: ?j < j ? P i

. Min{?-j-|-m-x,j-=-m-y,j-=-1}, Totally-Balanced-Dissimilarity-Approximation Data: A dissimilarity d on a n-set V Result: A totally balanced dissimilarity d on V, vol.6

, Approximate d into a chordal dissimilarity d admitting v 1 <

, (d )) defined by d (x, y) = min{? j | M x,j = M y,j = 1}

G. Alberti, Making Sense of Contingency Tables in Archaeology: the Aid of Correspondence Analysis to Intra-Site Activity Areas Research, Journal of Data Science, vol.11, pp.501-536, 2013.

R. P. Anstee, Hypergraphs with no special cycles, Combinatorica, vol.3, pp.141-146, 1983.

R. P. Antsee and M. Farber, Characterizations of totally balanced matrices, Journal of Algorithms, vol.5, pp.215-230, 1984.

H. Bandelt and A. W. Dress, Weak Hierarchies Associated with Similarity Measures -an Additive Clustering Technique, Bulletin of Mathematical Biology, vol.51, pp.133-166, 1989.

B. Barthélemy, Binary Clustering, Journal of Discrete Applied Mathematics, vol.156, pp.1237-1250, 2008.

P. Bertrand, Set Systems and Dissimilarities, European Journal of Combinatorics, vol.21, pp.727-743, 2000.

P. Bertrand and J. Diatta, Weak Hierarchies: A Central Clustering Structure Clusters, 2014.

F. Brucker, From hypertrees to Arboreal Quasi-ultrametrics, vol.147, pp.3-26, 2005.

F. Brucker and A. Gély, Parsimonious cluster systems, Advances in Data Analysis and Classification, vol.3, pp.189-204, 2009.

F. Brucker and A. Gély, Crown-free Lattices and Their Related Graphs, Order, vol.28, pp.443-454, 2010.

F. Brucker and . P. Préa, Totally Balanced Formal Concept Representations, Proceedings of ICFCA 215, pp.169-182, 2015.

J. Diatta, B. Fichet, Y. Diday, M. Lechevallier, P. Schader et al., From Asprejan Hierarchies and Bandelt-Dress Weak-hierarchies to Quasi-hierarchies, New Approaches in Classification and Data Analysis, pp.111-118, 1994.

G. A. Dirac, On rigid circuit graphs, Abh. Math. Sem. Univ. Hamburg, vol.25, pp.71-76, 1961.

M. Farber, Characterizations of strongly chordal graphs, vol.43, pp.173-189, 1983.

J. Lehel, A Characterization of Totally Balanced Hypergraphs, vol.57, pp.59-65, 1985.

L. Lovasz, Beiträge zur Graphentheorie, pp.99-106, 1968.

A. Lubiw, Doubly lexical Orderings of Matrices, SIAM J. on Computing, vol.16, pp.854-879, 1987.

J. Spinrad, Doubly lexical Ordering of Dense 0-1 Matrices, Information Processing Letters, vol.45, pp.229-235, 1993.

J. Spinrad, Efficient Graph representations American Mathematical Society, 2003.