Simon Blanke

@smilingprogrammer Could you elaborate on this? How would this line search look like? My original goal of this issue was to implement the calculation of the angle close to how...

> The HillClimbOptmizer as used in the Powells Method seems to offer no deterministic guarantee of converging to the true minimum in my new implementation. Good observation! The neighbour selection...

> You can find the implementation I am looking forward this. Could you open a PR for your changes?

Hello @AnonymityGithub, GFO is **not** designed to be used as an optimizer in deep-learning frameworks (like pytorch or tensorflow), because its optimization-algorithms work without utilizing the gradient (or 1. derivative)...

> However, after the emergence of LLMS, gradient-based optimization methods require larger memory, so gradient-free methods have emerged to optimize the model ... Intriguing! I was not aware of this....

I did some reading on this topic and worked out a concept, how to apply GFO to a pytorch model. As a first step I would like to apply GFO...

So, I did some tinkering with pytorch and managed to create a very simple example, how GFO could be used as an custom optimizer. Disclaimer: I don't use pytorch on...

As mentioned before I identified two reasons for the bad result when using GFO as a pytorch optimizer so far. I am currently working on the slow performance of GFO...

Hello @mosheduminer, welcome to this project. > pos_best consists of integers between 0 and 19 you are correct! The values in the search-space are converted into integers on optimizer level....

Hello @mosheduminer, you are correct about the maximization in GFO. I hesitated to update the example, because I thought you might want to open a PR. This way your contribution...