Scalable Graph Hashing with Feature Transformation Mode and Leoming Sequential Learning Algorithm Algorithm 1 Sequential learning algorithm for SGH Input:Feature vectors X E R";code length c:number of kernel bases mn. Output:Weight matrix W E Rexm Procedure Construct P(X)and Q(X)according to (3): Construct K(X)based on the kernel bases,which are m points randomly selected from X: Ao=K(X)P(X)Q(X)K(X): A1=CAo: Z=K(X)K(X)+Ia: for=1-cdo Solve the following generalized eigenvalue problem AW=AZW: U=K(X)Tsgn(K(X)w:)]K(X)sgn(K(X)w): A+1=A:-U: end for Ao=Ac+1 Randomly permutate {1,2....c}to generate a random index set M: fort=1 cdo t=M(t): Ao=Ao+K(X)Tsgn(K(X)wi)sgn(K(X)w)K(X): Solve the following generalized eigenvalue problem Aov =AZv: Update wv Ao =Ao-K(X)Tsgn(K(X)wi)sgn(K(X)wi)K(X): end for 三4元重)双0 Li (http://cs.nju.edu.cn/lwj) Learning to Hash LAMDA,CS.NJU 29 /43Scalable Graph Hashing with Feature Transformation Model and Learning Sequential Learning Algorithm Li (http://cs.nju.edu.cn/lwj) Learning to Hash LAMDA, CS, NJU 29 / 43