First, start from a blog post (with working code!) and then share it on https://www.reddit.com/r/MachineLearning/ https://www.reddit.com/r/MachineLearning/. Vi

First, start from a blog post (with working code!) and then share it on https://www.reddit.com/r/MachineLearning/ https://www.reddit.com/r/MachineLearning/. Vide: https://www.reddit.com/r/MachineLearning/comments/756xt2/p_experiments_with_a_new_kind_of_convolution/ https://www.reddit.com/r/MachineLearning/comments/756xt2/p_e... Academic journals are not a good starting point for pushing your findings, unless you are already very familiar with the field (publications, style of writing, etc). It does not suffice to have a new idea. And very easily you can get discouraged, for no good reason.