Cascade failure

A Cascade Failure occurs when a 'learning' program is corrupted, either by repeated application to a task it is ill suited to perform or by external influence such as damage to the hardware used to implement it.

In neural net systems where most or all sub-programs interact, this can lead to corruption of surrounding algorithms and routines as they attempt to use data generated by the faulty application to perform a task of their own. This may compound the damage to the original sub-program, as neural nets tend to be referential in nature.

If no limiting action is taken by the system's 'manager' programs and the process is allowed to continue unchecked, it may lead to partial or total corruption of the entire system.

For example: Program A attempts to analyse data block 1. The data is faulty but the 'answer' looks OK to Program A because it has never encountered such an anomalous result before and so doesn't recognise it as such.

Program B takes the answer and performs a more advanced operation on it. Again the outcome looks OK, so an 'OK' message is sent to Program A. Program A, since it is a learning program, moves that method of solving this type of problem up it's hierachical structure. Next time it will waste less time before trying that method. Every time a similar set of data appears, the scenario repeats until Program A's default method is the 'faulty' one and it will cease to ever obtain a valid resolution of any data set.

Program B, if not checked by a routine higher up the chain, will be similarly affected. And so on for C, D, E....

Normally, of course, at some point in this process, the operating system will notice that the data coming up from it's analytical agents is obviously faulty and will terminate the proceedure. A well designed system will also drop the priority of the methods used to process this type of data back to their original settings, which may slow down future operation but stabilises the system.

Of course, when you have the scenario of an android getting scared, for example, the data may be so foreign to the system's experience that the 'manager' never steps in to rectify the problem until it's far to late to stop the process and the management systems themselves are beginning to be affected. This is Cascade Failure. 'A' malfunctions, which damages B, which damages C.....

This article or section needs citationsThis page or section does not adequately cite one or more of its sources, and needs attention. If you can provide references from valid resource material, feel free to edit it to add a citation or correct the data.

This article is unformattedThis page requires formatting – possibly including wikification, and adherence to the Manual of Style – and has been identified as needing attention. Please feel free to edit this page to assist with this task.