Personal Perspective

Investigation of the Latent Space of Stock Market Patterns with Genetic Programming

We suggest a method of inspecting the latent space induced by matrix factorization. Given a relation matrix, we can apply standard techniques such as non-negative matrix factorization to extract low dimensional latent space as vector representations. While the vector representation of the latent space is useful, it is not intuitive and interpretable. We show that one can use additional information, apart from the relation matrix, together with genetic programming to gain insights of the latent space by constructing tree representations that correspond to vector representations. The ability to find a tree representation corresponding to an arbitrary latent vector representation allows us to better understand the underlying latent structure. Applying the method in the context of a stock market, we show that it is possible to recover the tree representation of technical patterns from a relation matrix. Leveraging the properties of the vector representations, we are able to find patterns that correspond to cluster centers of technical patterns. We further investigate the geometry of the latent space.

Here, we investigate the use of genetic programming to interpret latent spaces induced by low rank matrix factorization.