‘Backronym’ May Help To Generate Ideas in Machine Learning by Visualizing Hundreds of Research Papers Together

Source: https://backronym.xyz/graph.html

Every year hundreds of papers are published on Machine learning/Deep Learning in various journals and conferences. Most of them are based on a single method to solve a problem while some might have methods cross-referencing. The current methods and models are limited in number when we restrict ourselves to selected research papers while some research papers do cross-reference other methods which are quite helpful and have a lot more application than alone. The question is how and where to find and learn different methods to solve a problem and learn about models from research papers.

Arip Asadulaev has built “Backronym” tool to showcase hundreds of research papers together by cross-referencing research papers and methods. This tool is in early stage but still can help in finding various new and old methods related to each other for specific models.

For example GAN[1] consist of Generator(GEN), Discriminator (DIS), Adversarial Autoencoder (AAE)[2] based on Autoencoder (AE)[3] and DIS, . Every component is a separated node in the graph, so for AAE we will have an edge to AE and DIS.

AdvertisementCoursera Plus banner featuring Johns Hopkins University, Google, and University of Michigan courses highlighting data science career-advancing content

In the graph each research paper is presented as one single method, for example, Autoencoder (AE), or separately, Autoencoder (AE) -> Encoder (ENCDR), Decoder (DCDR), That is, one model can consist of several elements that can be used separately.

The graph/visualization is created using the 3d-force-graph javascript library: https://github.com/vasturiano/3d-force-graph


Read The Medium Article For Details

Website: https://backronym.xyz/graph.html

Paper: https://arxiv.org/pdf/1908.01874v2.pdf



Please enter your comment!
Please enter your name here

This site uses Akismet to reduce spam. Learn how your comment data is processed.