Title : 
Groups and information inequalities in 5 variables
         
        
            Author : 
Markin, Nadya ; Thomas, Edward ; Oggier, Frederique
         
        
            Author_Institution : 
Div. of Math. Sci., Nanyang Technol. Univ., Singapore, Singapore
         
        
        
        
        
        
            Abstract : 
Linear rank inequalities in 4 subspaces are characterized by Shannon-type inequalities and the Ingleton inequality in 4 random variables. Examples of random variables violating these inequalities have been found using finite groups, and are of interest for their applications in nonlinear network coding [1]. In particular, it is known that the symmetric group S5 provides the first instance of a group, which gives rise to random variables that violate the Ingleton inequality. In the present paper, we use group theoretic methods to construct random variables which violate linear rank inequalities in 5 random variables. In this case, linear rank inequalities are fully characterized [8] using Shannon-type inequalities together with 4 Ingleton inequalities and 24 additional new inequalities. We show that finite groups which do not produce violators of the Ingleton inequality in 4 random variables will also not violate the Ingleton inequalities for 5 random variables. We then focus on 2 of the 24 additional inequalities in 5 random variables and formulate conditions for finite groups which help us eliminate those groups that obey the 2 inequalities. In particular, we show that groups of order pq, where p; q are prime, always satisfy them, and exhibit the first violator, which is the symmetric group S4.
         
        
            Keywords : 
group theory; information theory; network coding; Ingleton inequality; Shannon-type inequalities; finite groups; group theoretic methods; information inequalities; linear rank inequalities; nonlinear network coding; random variables; symmetric group; Cramer-Rao bounds; Entropy; Joints; Mutual information; Network coding; Random variables; Vectors;
         
        
        
        
            Conference_Titel : 
Communication, Control, and Computing (Allerton), 2013 51st Annual Allerton Conference on
         
        
            Conference_Location : 
Monticello, IL
         
        
            Print_ISBN : 
978-1-4799-3409-6
         
        
        
            DOI : 
10.1109/Allerton.2013.6736607