Skip to content

Latest commit

 

History

History
25 lines (20 loc) · 1.37 KB

README.md

File metadata and controls

25 lines (20 loc) · 1.37 KB

memanalyzer

Code for sampling and analysis of computer program execution data (Linux)

For now it reads target process (PID) memory and tracks changes in memory use.

The final idea is to also collect kernel execution information and make computer more "self-aware" by turning deep learning algorithms to analyse (also) internal state which should then make a model where own internal state will be part of mathematical models that are used to understand world.

The preprocessing challenge is that data vectors are large (10^7 dimensions O(10 MB)) and only 3% of memory change in ~20 Hz sampling frequency. This means large part of data needs to be ignored. For this calculating PCA reduction to keep 90% of variance without computing covariance matrix is needed.

Additionally, during program execution there are only small active spots that keep changing so the area of memory change probably keeps changing rapidly and faster than our maximum speed sampling frequency (~ 20 times a second).

After reducting dimension of variables to something like 10^4 we can then maybe compute cross correlation matrix Cyx when data representation changes (memory allocations/memory maps, shift to new set of active variables) and compute linear transformation y = Ax to keep representation of active variables (we should try try to follow computation like continuous thought process in human brain).