[akhil999in]

Can we simulate a computer that modifies itself to statistical feedback ?

We define Statistical properties of the output of a computer.
For instance, the variation of ratio of Zeros and Ones in its Output
Stream.

Then we define feedback in terms of Statistical properties of output.

We Ask the computer to modify its circuits accordingly.

For this we have to Set a Target in the same Statistical terms, e.g. it could be
a number, Say 1.863 = Value of Ratio of Zeros and Ones. The Computer must
modify its circuits so that it can achieve the Target within some Criterion of
degree of Success.

Or for another even simpler case, let the computer be Asked to behave like a
kaleidoscope, i.e. random but Meaningful rearrangements of its Internal
Structure.

Is it possible to simulate the Computer and read off its Output and also its
momentary Circuit Design from time to time? Has it been done ? Are there any
useful generalizations of the observed behaviour and any theoretical explanation ?

[tickle]

Why would anyone want to 'simulate' a computer? Possibly you have used the wrong word to convey your intent. Computers are the ultimate in 'simulation'.