Executable Specification of Dose Escalation Oncology Protocol now on Arxiv

1

I still have to read the paper, but @davidcnorrismd and the computer scientist Markus Triska, have published a paper demonstrating the properties of a clinical trial design using one of my favorite computing languages – Prolog. This is a wonderful intersection of medicine, statistics, and computing science.

See also:

https://discourse.datamethods.org/t/continuity-of-your-work-do-you-have-a-protege/4524/4

2 Likes
2

Many thanks for highlighting this, Robert! I’m glad to know we share an interest in Prolog, too! I’d be happy to do an AMA, for anyone interested in this paper. I am in fact hard at work these days on a category-theoretic formulation of dose-escalation trial designs aiming:

  1. to exhibit certain ethical shortcomings of these designs as [what is known in CT as] a universal property, and
  2. to embed dose escalation within a larger class of dose-titration designs that transcend these shortcomings.

Prolog remains my experimental tool of choice for these investigations.

1 Like
3

Speaking of category theory, there is an interesting dissertation that puts statistical models in a category theoretic context that will make it easier to compare and contrast traditional statistical models with machine learning. I am wondering how long it will take someone to use this framework to address some of the questions @f2harrell has asked in some of his ML vs Stats posts.

The video is likely to be abstract nonsense to those not familiar with terminology from category theory (I’m certainly no expert), but I post it for completeness.

Related thread:
https://discourse.datamethods.org/t/relationship-between-statistical-models-and-machine-learning-algorithms/7063?u=r_cubed

2 Likes