Introduction and Navigation

The introduction to double descent - the particular phenomenon where current dogma faces reality as it sees fit.

This is the main central page for double descent research, including documentations, current work, updates, many and much about what is going on with the research, and just simply the dump we pull stuffs here and showcase our research progress (with a crowbar, or a progress bar if someone is good on it). Insofar, it will also include thoughts and debates on such matter, aside from actual contents in manuscripts and other pages, mainly the wiki page, which has been up for a long time now.

TipConnection to theoretical machine learning

Double descent research is part of the series of theoretical machine learning, and as a subseries also of the theoretical artificial intelligence course. The status of double descent is particularly to diagnose specific foundational problems, and thus it remains contingent to the wider scope. Identification of the majority of double descent occurrence and investigation lies in statistical machine learning, and not computational machine learning, respectively.

Information

Here, we can dub the project or rather classification thereof for this section, as being a theoretical inquiry module that is conducted to serve a larger picture there be, of the theoretical modelling principles system of artificial intelligence. However, for such to flow, it requires us to introduce the central problem of what to be said of double descent.

Research goal

The current goal of the research group by definition, is to follow the underlying analytical results, stated:

Research resources

As far as it is considered, our progresses are in the end of phase II, i.e. the final phase of the core theoretical discourse, analysis and dissection. Phase I is the probing, analytical fundamentals gathering, and review performing, basically to build the relevant resources and materials needed for a serious critique and expansion thereof. Insofar, we have the following.

Principal guideline

The Universal Guideline on Artificial Automata - still under construction, is the foremost document itself. The name is again, Amane Fujimiya or Bui Gia Khanh (as I use two names), and is more about theoretical attempts on AI theory in its entirety. So far, this link will be the one that gives you the last available rendered manuscript of the book (note that it is on the master branch). The other documents, the one accessible of physical learning theory, is also available.

Knowledge and documentation, laboratory wiki

There are many of such resources, which is realized after the fact and observation that the discovery and research about double descent, is rather profound and can be said quite so as impressively complex, with many logistics to be fulfilled. As a result of said resolution, the decision is then used to build up the knowledge and documentation logistics, making up the complete or rather aimed for of a comprehensive system of information, documentation, frontier discussions and ideas, and tracks. Most often would be useful however, to check out for the expository papers, knowledge base and laboratory-specific wiki, in connection to the wider RHINELAB, though this would for now take precedences of only the domain of theoretical machine learning.

Core manuscript

The Main manuscript - constructed on This Particular Repository (TPR) of which holds from the first to the last manuscript. Current active manuscript in total is the Draft folder itself document. Alongside that, this paper would form chapter 6 of the manuscript itself for analysis, because the issue with model complexity is a rather interesting and very much understudied area.

The manuscript of the Theoretical Learning manuscript of this project can be found in here of its latest version.

Main manuscripts, statistical physics - is a separation from the main one. It is mostly focused on interpretation of learning theory in statistical physics, dissection of such, and so on.

Experiments and setup

For experimental setup, code and implementations, we have a few one, most of their results are given in the major manuscript.

  • The code repository includes this one on GitHub, with its copy and extension on GitLab. The link will be routed later. We note that this is the canonical repository for testing and crude experimental sessions. Some of them would be messy, and in the meantime, a second repository is coming up as expansion.
  • A separated library for testing and identifying double descent, and structural analysis. Currently, it is in private build-up, with many features and analytical toolings missing.