I have come up with a new function fitting method, where we fit a function to a given data. I have described the mathematics of it, and proved relevant theorems to show how it works. The sole importance of this method is its applications to machine learning. If I go to mathematicians, they are saying that the theorems are right, but the mathematics of it is nothing unusual but expected. They don’t readily know much about Machine Learning, and have no inclination to know. So its difficult to impress a math journal editor for a publication acceptance. So the mathematicians are advising to go to machine learning experts. On the other hand, If I go to machine learning experts, they are reluctant to comment, as they don’t readily understand the relevant math (unless they take some time and refer to a few books). Moreover the concept is a bit counter intuitive to the latest beliefs in machine learning world, where almost everyone believes that the ML problems are to be solved in very high dimensions, and most of the successful tools like deep convolutional neural networks or the traditional kernel methods or the graph based methods, are designed, keeping high dimensions in mind and over the belief that the ML problems can only be solved in very high dimensions. My methods calls for solving in as low dimensions as possible, requiring traditional domain knowledge based feature representation combined with dimensionality reduction tools, as pre-processors to reduce dimensions. So if I talk about my method to ML people, they might say that ML problems are best solved in very high dimensions, so my method being virtually impractical for very high dimensions, they deem it useless for machine learning.

I am able to apply my method and and demonstrate, for solving a few ML datasets of the likes of IRIS. I am also able to show the inner workings of my method through visualizations on simulated datasets in 2 dimensions, just for sake of illustrations.

I need better workstations and some time and funding to apply and solve harder ML problems, for which I need some support and funding, which is possible only if someone buys my idea and sponsor, as a form of startup. My strategy is to first publish this mathematical method of function fitting, in a math journal, so that it gets some authenticity and help me get some serious attention from ML experts for providing labs/infrastructure or attract venture capitalists for a startup.

Appreciate some suggestions whether my strategy is good idea. If so, what are some math journals I can target for this purpose. I don’t expect to go to mathematicians and say that I have done something incredible, but I just want to garner enough interest to get published in a descent journal, so that it will be easy for me to gather attention from ML world.