Re:One parameter is always enough [pdf]
by tomtimtall on Saturday May 26, 2018 @02:29PM
(#17163559)
> that “parameter counting” fails as a measure of model complexity when the class of models under consideration is only slightly broad.
This seems plainly false. They only show that for this one specific function with one parameter determined to several million digits. That would not by any means be “only slightly broad”.
They are essentially just restating that any dataset can be expressed as a single binary number therefore it can be “fitted” by a function that has a completely covering map between integers to integers. While I find it an interesting though not surprising element that they did it with the logistic map, their claimed purpose and conclusion are really far fetched.
This seems plainly false. They only show that for this one specific function with one parameter determined to several million digits. That would not by any means be “only slightly broad”.
They are essentially just restating that any dataset can be expressed as a single binary number therefore it can be “fitted” by a function that has a completely covering map between integers to integers. While I find it an interesting though not surprising element that they did it with the logistic map, their claimed purpose and conclusion are really far fetched.