__lm__

There is still a difference. Previously, if the review was low quality and superficial, it was also quite short and easy to answer. Now I receive fully LLM-written reviews of 2-3 pages (or more) of superficial comments asking for tons of additional text and experiments that are almost always outside the scope of the work. Answering such a review requires a lot of work with absolutely no gain.

We have some older Power9 with NVIDIA V100. Nvidia drivers stopped a while ago (before V100 were considered “old” also for x86_64).

The main problem is software support by most machine learning libraries. While it is normal to have to compile many things from source when the binary is not provided even by some third party, in some cases the software will not compile on ppc64le and surely is not tested to work.

Yes, unfortunately not open source. I tried Open Genera with the VLM port for Linux about ~8-10 years ago. The tutorials and the process got a lot better a few years laters. This should be an official way to run it (with all the associated additional first-party programs).

From an historical point of view it is interesting to see what “could have been” and what was lost.

Even recent works might be both unscanned and untranslated.

I experienced this when the only way to access a 70 years old scientific paper written in Italian was by finding a copy in a library and scanning it. There are not many existing copies of the journal left, in a few decades it might have been lost forever.

Singularity (now called Apptainer - https://apptainer.org/) is a container system generally used in HPC environments that has some nice features, like having your container as single .sif file, automatically mounting your home directory, using the same user inside and outside the container, and the ability to import from the docker registry. The approach is less about isolation and more about “packaging up” an entire environment in a easy-to-use way.

Useful to explain non-determinism to students. I saw a similar idea before at https://github.com/aeporreca/nondeterminism which uses fork() to (inefficiently) explore all possible guesses concurrently

When Sun Microsystems was still existing they made a (quite funny) video putting 12 usb sticks in a ZFS pool with RAID: https://constantin.glez.de/2011/01/24/how-save-world-zfs-and...

NEAT and neuroevolution in general are interesting approaches. I also suggest to check techniques like DENSER [1] that can be used to evolve deep networks (by using the evolutionary part on the network structure and not on the weights).

Genetic Programming (GP), however, has not evolved to NEAT (which itself is not very recent, being published in 2002) but simply neuroevolution has become one of the topics that are part of evolutionary computation (EC). For example, one of the largest yearly conferences on evolutionary computation (GECCO) [2] was just last month with both neuroevolution and GP tracks. It is however true that the success of neural techniques had an effect on the community, some effects are the discussion of the role of EC and, for example, more space given to hybrid works (see, for example, the joint track on evolutionary machine learning [3] inside the evostar event).

Related to the original post, a place where some recent research on GP can be found are the proceedings of GECCO (GP track), EuroGP (part of evostar), PPSN (Parallel Problem Solving from Nature), and IEEE CEC (IEEE Congress on Evolutionary Computation) and journals like Genetic Programming and Evolvable Machine (GPEM), Swarm and Evolutionary Computation (SWEVO), and IEEE Transactions on Evolutionary Computation (IEEE TEVC). The list is not exhaustive, but those are some well-known venues.

For a less "daunting" starting point, some recent techniques are being added to the SRBench benchmark suite [4], with links to both the code and the paper describing the technique.

[1] Assunção, F., Lourenço, N., Machado, P., & Ribeiro, B. (2019, March). Fast denser: Efficient deep neuroevolution. In european conference on genetic programming (pp. 197-212). Cham: Springer International Publishing.

[2] https://gecco-2023.sigevo.org/HomePage

[3] https://www.evostar.org/2024/eml/

[4] https://github.com/cavalab/srbench

The first one is available here: http://regex.inginf.units.it/

It uses genetic programming to build the regular expression.

That is exactly what I would also like to do. The HN comments are usually as interesting as the the link in the post. For example people point out other resources or the author of the linked post comments with more details.

I still have no clear solution to this, I currently use org mode but 1) it does not work well on mobile and 2) even on the desktop I should find the time to automate it more (with automatic archiving, better templates captures, etc). While 2) can be solved, 1) is still a problem.

That looks interesting!

Does it also have a way to deal with sets of "related" links? For example I might want to save and archive a webpage and its discussion on HN together or a collection of multiple blog posts as single entity. Currently I use org-mode but, even with org-capture, it is not frictionless and the mobile integration is lacking.

In Italy it is one of the three metrics used. The others are the number of published papers and the h-index.

This is the table of thresholds for associate professorship ("II Fascia") and full professorship ("I Fascia"): http://abilitazione.miur.it/public/documenti/2018/Tabelle_Va... "Numero articoli" is "number of papers", "Numero citazioni" is "number of citations", and "Indice H" is "h-index". The thresholds are different for each research area (e.g., "INFORMATICA" is "computer science").

The topic is actually a little bit more complex, since looking at the metrics is only one step of the process.

It was edited, before it was 80 H/s.

Wasn’t the speed of the IBM 1401 only one hash every 80 seconds (1/80 H/s)?

My Raspberry Pi 3 works as a CUPS print server connected to a laser printer, as a pi-hole DNS server to filter ads, and as a ssh entry point (with dynamic DNS). I would use a Pi zero, but Ghostscript is not too happy when printing large documents on a single-core processor with 512 MB of ram. I still have to find the time to set-up a backup server on it (and decide which software to use).

Good job! Homebrew is, at least for me, essential for using macOS as a true *nix, and most of the changes (implemented or proposed) seem good (e.g., the ability to automatically run cleanup)

I am a little bit worried for the removal of all options that is planned for the 2.0.0 version: I have some formulae installed from source (with additional options) and I like the flexibility that they give. Hopefully the migration to an "options-free" Homebrew will be smooth and some alternative when options are needed (not requiring a lot of third-party taps) will emerge.

The fact that there is no standard set of benchmarks was recognised some years ago:

https://cs.gmu.edu/~sean/papers/gecco12benchmarks3.pdf

http://www.cs.put.poznan.pl/wjaskowski/archives/wp-publicati...

While I think that the situation is improving, we are still far from a common set of benchmarks used by the entire GP community.

On Mac I use the emacs-mac port by Mitsuharu Yamamoto: https://bitbucket.org/mituharu/emacs-mac/overview which, in my opinion, works well. It can also be installed via brew (from source) or brew cask: https://github.com/railwaycat/homebrew-emacsmacport