Is there another way to see of the whole market beyond composite indexes and statistical features? How to represent the market beyond top-down categories? It is all about “representation” and “resemblance”. Lets see the whole market and its fluctuations as a time varying random field. Then the question that I am asking is what would be the best structure of this random field?
Assume each symbol (in the NYSE market for example) is a pixel in an image. Then the question is which pixel should be the neighbor of which other pixels to finally give us a better image of the market? Like when you see an image of a face and you see the face is smily, but you don’t care about individual pixels of the image that much.
This is the main idea of my recent project that I call an image of the market. It can be an image of the city, but definitely different than what Kevin Lynch greatly did in 1960.
I will mainly use ideas from image processing, spatiotemporal dynamics, Markov random fields, and high dimensional time-series forecasting.