How can noise be good?

Question

It seems like (controlled) noise can sometimes be good: dithering in an image, or a softer sound in audio, or white noise alone. Here's a paper on noise improving detectability. My intuition also says that noise in a financial trading algorithm could act as a sort of hedge.

I have few examples — only a vague philosophical instinct. Is there any general-purpose literature or theory on why and how noise can be good? Or other/better examples?

because it reduces distortion http://dsp.stackexchange.com/a/19451/29 — endolith, Mar 23 '15 at 23:12

score 4 · Accepted Answer · edited Mar 23 '15 at 21:36

4

Another example is setting up A/D converters - usually the input is set so that the noise just trips the least significant bit (LSB) of the converter. The noise can also be added manually - this process is also referred to as Dithering. The white papers at various A/D manufacturers can be quite useful e.g. Linear Technologies, Analog Devices and Texas Instruments.

Another term you may wish to look under is Stochastic Resonance.

edited Mar 23 '15 at 21:36

lennon310

3,590
19
24
27

answered Mar 23 '15 at 17:37

David

2,861
11
13

Wow, "stochastic resonance" (http://en.wikipedia.org/wiki/Stochastic_resonance) is exactly what I was looking for — I was actually thinking about resonance in the back of my mind. So your answer resonated very strongly with me thanks to a little creative leap you made, which seems like a great example of stochastic resonance! – Toph Mar 23 '15 at 19:35
@Toph In the paper by Kay (the one you linked to) - he does mention Stochastic Resonance, so you may find additional references at the end of his paper. – David Mar 23 '15 at 19:52

How can noise be good?

1 Answers1