thanks for your reply.
the purpose for this question is to do transfer function expression approximation for symbolic circuit analysis. such as description in this link
because there may be many , such as hundreds of sub-expression in each plus items in a transfer function, like this
H=( ( (exprA1+exprA2+…) * (exprB1+exprB2+…) )/( (exprC1+exprC2+…)* (exprD1+exprD2+…) )
In a plus item, such as (exprA1+exprA2+…), each sub-expression is a product of circuit parameters, the number of sub-expressions may large (hundreds of), but in a circuit typical design point, the value of most of these sub-exrpessions is much smaller than the others, they can be droped.
For example, if at a circuit’s typical design point, eval each sub-expression, we have
Abs(exprA1) >> abs(exprA2), abs(exprA3), …, abs(exprAN),
Then the plus item (exprA1+exprA2+…) can be approximately be (exprA1), so the transfer function is more simple for a circuit designer.
To do symbolic approximation, I need to find all the position in DAG of these sub-expression, estimate each subexpression at design point in the transfer function, if its value is smaller than a threshold, this sub-expression can be substituted with 0.