mine discret pattern binari matrix factor bao hong shen shuiwang ji jiep ye mine discret pattern binari data import subsampl compress cluster rank binari matrix approxim identifi domin pattern data preserv discret properti approxim data minimum set inconsist entri mismatch binari data approxim matrix due hard problem previou account problem emploi heurist result approxim optim paper show rank binari matrix approxim reformul integ linear program ilp ilp formul computation expens small size matric propos linear program lp relax shown achiev guarante approxim error bound extend propos formul regular techniqu commonli emploi address overfit lp formul restrict medium size matric due larg number variabl involv larg matric interestingli show propos approxim formul transform instanc minimum cut problem solv effici find maximum flow empir studi show effici propos algorithm base maximum flow result confirm establish theoret bound 