• + 1 comment
    In order to solve this problem, first you need to know how to use dp to solve permuation problem.
    ex: N=3, dp[i][j], i means total numbers, j means last element. (j<=i)
    
    dp[1][1] = 1			//1
    
    dp[2][1] = sum(dp[1]);		//(1) 1	=> 2 1 (replace 1 with 2)
    dp[2][2] = sum(dp[1]);		//1 2
    
    dp[3][1] = sum(dp[2]);		//2 (1) 1 => 2 3 1 (replace 1 with 3)
    				//(1) 2 1 => 3 2 1 (replace 1 with 3)
    								
    dp[3][2] = sum(dp[2]);		//(2) 1 2 => 3 1 2 (replace 2 with 3)
    				//1 (2) 2 => 1 3 2 (replace 2 with 3)
    								
    dp[3][3] = sum(dp[2]);		//2 1 3
    				//1 2 3
    								
    total permutation is sum(dp[3]) = 6.
    
    
    Now we need to think about constraints a and b. 
    We can define as following:
    a(i)=1 as a has ith position element
    b(i)=1 as b has ith position element.
    
    There are total three situations for dp[i][j]:
    
    if a(i) == 1 or b(i-1) == 1		(include only bigger sets)
    	sum(dp[i-1][j] + dp[i-1][j+1] ... dp[i-1][i-1])
    else if b(i) == 1 or a(i-1) == 1	(include only smaller sets)
    	sum(dp[i-1][1] + dp[i-1][2] ... dp[i-1][j-1])
    else
    	sum(dp[i-1])
    
    
    Ex: N = 4 a1 = 2 and b1 = 3. The 5 permutations of {1,2,3,4} that satisfy the condition are
    2 1 4 3
    3 2 4 1
    4 2 3 1
    3 1 4 2
    4 1 3 2  
    
    
    dp[1][1] = 1			//1
    --------------------------------------------------------------
    a(2) = 1
    dp[2][2] = 0	    		//j=2, no bigger sets
    dp[2][1] = dp[1][1]		//(1) 1 => 2 1
    ---------------------------------------------------------------
    b(3) = 1
    dp[3][3] = dp[2][2]		//0
    		   + dp[2][1]  	//2 1 3
    		 
    dp[3][2] = dp[2][1]		//(2) 1 2 => 3 1 2
    
    dp[3][1] = 0			//j=1, no smaller sets
    ---------------------------------------------------------------
    b(3) = 1
    dp[4][4] = 0			//j=4, no bigger sets
    
    dp[4][3] = dp[3][3]		//2 1 (3) 3 => 2 1 4 3
    
    dp[4][2] = dp[3][3]		//(2) 1 3 2 => 4 1 3 2
    		   +dp[3][2]	//3 1 (2) 2 => 3 1 4 2
    		   
    dp[4][1] = dp[3][3]		//2 (1) 3 1 => 2 4 3 1 => 4 2 3 1
    		   +dp[3][2]	//3 (1) 2 1 => 3 4 2 1 => 3 2 4 1
    		   +dp[3][1]	//0
    
    You may realize we can always change order of dp[i-1], because total permutation doesn't change.
    ex:   2 1 3, we have relationship 2 > 1 < 3
    after change to 2 4 3, we can always have same relationship 4 > 2 < 3.
    
    result sum(dp[4]) = 5