Comments on: The topology of dreams https://lievenlebruyn.github.io/neverendingbooks/the-topology-of-dreams/ Sat, 31 Aug 2024 11:07:38 +0000 hourly 1 https://wordpress.org/?v=6.6.1 By: lievenlb https://lievenlebruyn.github.io/neverendingbooks/the-topology-of-dreams/#comment-147 Fri, 03 Mar 2023 12:18:30 +0000 http://www.neverendingbooks.org/?p=10937#comment-147 In reply to javier.

Thanks Javier!

]]> By: javier https://lievenlebruyn.github.io/neverendingbooks/the-topology-of-dreams/#comment-146 Tue, 28 Feb 2023 11:07:27 +0000 http://www.neverendingbooks.org/?p=10937#comment-146 Hi Lieven! Hope everything is going well.
This was an interesting post! Perhaps you are already angling in this direction, but in case you are not, I think your remarks about the need for a topology would connect nicely with some of the ideas from Topological Data Analysis (see for instance https://www.ams.org/journals/bull/2009-46-02/S0273-0979-09-01249-X/S0273-0979-09-01249-X.pdf).
This was all the rage in the data analysis circles a few years back, the core idea being that when we observe data the values we see are not natural because they rely on us picking “coordinates” or “variables” to measure, and thus to understand the true nature of the data we should be looking at some kind of underlying topology. The field of TDA is full with papers on how to use reconstruction functors to create topological spaces/simplices that best describe an observed dataset. I haven’t seen anyone trying to apply it to language models, but then again I don’t follow developments in NLP/LLM very closely.

]]>