From 1ef0a91292ba0a8b2dc6cfa5b8fe1bb6a872d46a Mon Sep 17 00:00:00 2001 From: Tanya Strydom Date: Thu, 18 Jul 2024 10:21:20 +0100 Subject: [PATCH] =?UTF-8?q?=F0=9F=92=AC=20thoughts=20on=20the=20cost=20of?= =?UTF-8?q?=20prediction?= MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit --- 01_reconstruction_methods.qmd | 12 +++++++++++- 1 file changed, 11 insertions(+), 1 deletion(-) diff --git a/01_reconstruction_methods.qmd b/01_reconstruction_methods.qmd index afa2daa..0a4d9f7 100644 --- a/01_reconstruction_methods.qmd +++ b/01_reconstruction_methods.qmd @@ -28,7 +28,17 @@ bibliography: references.bib - Challenges we face (even in contemporary settings)? - - keep high level - I think the argument here should fall more in the data tradeoffs... + - keep high level - I think the argument here should fall more in the data trade offs... + +## The cost of prediction + +- different models need different data and/or are making some very opinionated assumptions + +- this is going to link back to the goal of prediction as well (desire to discover or describe) + +- which means need to make a cost-benefit analysis + +- especially if we want to think about how this intersects with model performance/benchmarking ## Understanding how networks are different