next up previous contents
Next: Information correlation functions Up: Information correlationcumulants, clusters, Previous: Clustering

Function hierarchies

Now, consider the case where we have clustered by support. The U(v)'s which appear all appear as a sum in an exponential when the representation tex2html_wrap_inline13545 is made. This yields the product representation tex2html_wrap_inline13547 for tex2html_wrap_inline13549. By extension to power series other than that for Exp, we may write any average of any function tex2html_wrap_inline13553 as tex2html_wrap_inline13555, where tex2html_wrap_inline13557 means take the sum of the functions arguments over arguments that are unique ordered subsets of tex2html_wrap_inline13559. Thus, tex2html_wrap_inline13561, etc. When we have a hierarchy of functions tex2html_wrap_inline13563, or when we generate such a hierarchy from a single function's series expansion, we can solve this system for the tex2html_wrap_inline13565.



David Wolf
Tue Mar 25 08:11:49 CST 1997