@mgorny@social.treehouse.systemsSometimes it makes sense to act smart rather than brute-force.
For example, when Intel makes another #MKL release and you get version like "2026.0.0", and you need to figure out the remaining "-n" suffix for the .deb packages. And you really don't want to start a Debian container to figure that out.
Well, you could just keep brute-forcing until you find the right number. Or you can figure out that the index URL is #Gentoo
@mgorny@social.treehouse.systemsWhy did Deutsche Bahn number the two connecting cars 628 and 928? Is it because they connect as 69?
@arXiv_csLG_bot@mastoxiv.pageOn the Generalization Behavior of Deep Residual Networks From a Dynamical System Perspective
Jinshu Huang, Mingfei Sun, Chunlin Wu
https://arxiv.org/abs/2602.20921 https://arxiv.org/pdf/2602.20921 https://arxiv.org/html/2602.20921
arXiv:2602.20921v1 Announce Type: new
Abstract: Deep neural networks (DNNs) have significantly advanced machine learning, with model depth playing a central role in their successes. The dynamical system modeling approach has recently emerged as a powerful framework, offering new mathematical insights into the structure and learning behavior of DNNs. In this work, we establish generalization error bounds for both discrete- and continuous-time residual networks (ResNets) by combining Rademacher complexity, flow maps of dynamical systems, and the convergence behavior of ResNets in the deep-layer limit. The resulting bounds are of order $O(1/\sqrt{S})$ with respect to the number of training samples $S$, and include a structure-dependent negative term, yielding depth-uniform and asymptotic generalization bounds under milder assumptions. These findings provide a unified understanding of generalization across both discrete- and continuous-time ResNets, helping to close the gap in both the order of sample complexity and assumptions between the discrete- and continuous-time settings.
toXiv_bot_toot
@markhburton@mstdn.socialAsk your MP to support a UK Digital Sovereignty Strategy https://action.openrightsgroup.org/protect-britain’s-digital-backbone-—-ask-your-mp-act
My MP won't sign EDMs (waste of public money he claims!) but …
@degrowthuk@mstdn.socialDebates on Degrowth: what drives us to keep growing?
Editors' note We are delighted to publish this article by Marga Mediavilla on a systems approach to degrowth. It serves to summarise a number of the issues in the Prospects for Degrowth series and also represents an echo of the methodology used in the report which was a major influence on degrowth thinking, the 1974 Limits to Growth report by Donella Meadows and colleagues.
@rene_mobile@infosec.exchange
@cdarwin@c.imCOVID may have killed significantly more people in the U.S. in the first two years of the pandemic than official records indicate, with as many as one overlooked death for every five recorded ones. That brings the total to nearly one million deaths just in 2020 and 2021.
That calculation comes from research published today in Science Advances that seeks to understand how many COVID deaths fell through the cracks of official reporting systems. The untallied cases show the burden of the …
@mgorny@social.treehouse.systemsDid you know that #PEP425 ("Compatibility Tags for Built Distributions") said:
> Why isn’t there a . in the Python version number?
>
> CPython has lasted 20 years without a 3-digit major release. This should continue for some time. Other implementations may use _ as a delimiter, since both - and . delimit the surrounding filename.
This didn't age well.
#Python
@degrowthuk@mstdn.socialYour weekend reading:
"We all know that our addiction to growth must stop because it is destroying the very foundations of our lives, but knowing that we must stop an addiction is one thing, and being able to do so is a different one. Capitalist growth is a complex and insidious dynamic that should be analysed through the lens of systems analysis and feedback in order to be addressed."
Debates on Degrowth: what drives us to keep growing?
@mgorny@social.treehouse.systemsWhen you're sitting in a different compartment than you had reservation for (why crowd with others when there are entirely free compartments?), and you're about to go to the toilet and leave your clothes, you need to remember the compartment's number.
So I'm walking and repeating "eight, eight, eight…" After a short while, I get distracted, and then I start wondering "what was that eight about?"
@arXiv_nlinCG_bot@mastoxiv.pageMeasuring the Computational Power of Finite Patches of Cellular Automata
Attila Egri-Nagy, Chrystopher L. Nehaniv
https://arxiv.org/abs/2604.14966 https://arxiv.org/pdf/2604.14966 https://arxiv.org/html/2604.14966
arXiv:2604.14966v1 Announce Type: new
Abstract: Computational power can be measured by assigning an algebraic structure to a computational device. Here, we convert a small patch of Conway's Game of Life into a transformation semigroup. The conversion captures not only time evolution but also interactive operations. In this way, the cellular automaton becomes directly programmable. Once this measurement is made, we apply hierarchical decompositions to the resulting algebraic object as a way of understanding it. These decompositions are based on a macro/micro-state division inspired by statistical mechanics. However, cellular automata have a large number of global states. Therefore, we focus on partitioning the state space and creating morphic images approximations that can serve as macro-level descriptions. The methods developed here are not limited to cellular automata; they apply more generally to discrete dynamical systems.
toXiv_bot_toot
@degrowthuk@mstdn.social
@primonatura@mstdn.social"More than 6m vapes and pods discarded weekly in UK despite single-use ban, study finds"
#UK #UnitedKingdom #Vapes
@hex@kolektiva.socialAs salty as I am about it, there's also another way to think about this. For anyone who still has connections to folks on the right (which is perhaps unlikely for anyone on this server, I digress), the cult that has consumed them thrives on isolation and grievance.
The words "you were right" have the potential to cut through the programming and open up an opportunity for reconnection. The modern conspiratorial cult of the Right has been built partially around people who were told they were wrong or were crazy. In the vast majority of cases, they were wrong and even when they were right they completely misunderstood why, but we'll skip that for now. Liberals making fun of them (even the times when they definitely earned it) has pushed them further and further into their ideological hole.
The thing about those words, "you were right," in this context is that the way they offer reconnection also requires them to take one little step of betraying their ideology to accept them. So they must choose between maintaining allegiance to a pedophile or finally getting to feel superior after years of living in an illusion of persecution.
Under the ideology of the Right, admitting one is wrong is a weakness. It is admitting defeat. They have to "own the libs" by saying things, things that they know aren't true, in order to feel dominant. But these things are often so absurd that they end up being made fun of, feeling even more weak and pathetic, reinforcing their fear and alienation.
Offering what they're looking for can offer a way out, but only if they're willing to start to recognize the thing they've supported for what it is.
And they were right about some things. They were right that Bill Gates was a terrible person. I've had plenty of liberals defend him based on his philanthropy washing, but he's awful and always has been. The Epstein links make that blatant. They intuitively recognized him and didn't trust him, even if they were wildly off base about *how and why* he shouldn't be trusted... Even if their correct mistrust was leveraged into one of the most destructive conspiracy theories ever (vaccine denial and COVID vaccine avoidance).
They were right about Bill Clinton. He was always shady as fuck. Sure, the people who attacked him at the time turned out to be even more shady but that's not the point right now. He was connected to Epstein and that was always creepy as fuck.
And the Epstein thing was an open secret that liberals ignored for a long time. It was seen as some weird thing that right wing nutjobs believed about the Clintons. But it was true. Not all of it, and there has always been an antisemitic element to the right wing interpretation or Epstein stuff, but his whole pedophile conspiracy was always kind of real.
The whole "Illuminati"/deep state thing is a vast oversimplification, an attempt to make comprehensible an incredibly complex set of interlocking and emergent behaviors. But Epstein did very much want to remake the world, to create a new world order, and he absolutely played a part in it.
The Right wing nutjobs talked about global authoritarianism, Blackhawks flying over American cities, masked men with guns disarming and executing legal gun owners in the streets. That's all happening right now.
The "FEMA concentration camps" are not actually that far off. ICE and FEMA are sister agencies, both under DHS. I'd be more than happy to call that one "close enough" in order to hear some MAGA admit that ICE is, in fact, building concentration camps.
There was always a huge millennialist element to these things. They tended to be connected to "the antichrist." It was absurd, especially for me as someone who no longer identifies as a Christian. But I'll even acquiess that to a degree. The "the number of the Beast" is 666. That's just the sum of the Hebrew spelling of "Nero." Revelations focuses a lot on Nero coming back to life after his death. His death that involved a head wound, thus the line from Revelation 13:3:
> And I saw one of his heads as if it had been mortally wounded, and his deadly wound was healed. And all the world marveled and followed the beast.
The parallels between Trump and Nero are easy to draw, and Trump's ear wound feels pretty on-the-nose for this. I don't believe in "prophecy" in this way. I think that there are patterns, and useful patterns can become encoded in beleif systems. But I will, again, happily call this one "close enough" for anyone on that side willing to also acknowledge it. I'm happy to meet on that common ground, because anyone who accepts it must recognize that their duty is to fight against it.
A lot of these correct nuggets are embedded in a framework of religious extremism and antisemitism. The vast majority of the beliefs holding these together are wildly wrong and incredibly toxic. But by giving some room to feel validated, listened to, understood, can give some room to admit things that were wrong.
Cult de-programming starts with an opening. People have to talk through their own thoughts, hear their own inconsistencies. Guiding questions can help them untangle these things for themselves. And it all starts by having enough room to feel safe, to not feel cornered, to not feel stupid. Admitting mistakes means being vulnerable, and the MAGA cult is built on fear. It's built on exploiting vulnerability and locking it away.
De-programming takes a long time. It's not easy. It takes patience. But every person who comes out does so with a powerful perspective, a deep understanding, that can be turned back against it. The best people at getting people out of cults are former members. Some of the most dedicated antifa are former fascists who understood their mistakes and dedicate their lives to fixing them.
@degrowthuk@mstdn.socialDebates on Degrowth: what drives us to keep growing?
By @MargaMediavulla
"In this article I propose a Systemic Analysis of growth and use some System Dynamics diagrams to analyse the problem."
It's an excellent piece.
https://degrowthuk.org/2026/04/16/deb…
@arXiv_csOS_bot@mastoxiv.pageBounded Local Generator Classes for Deterministic State Evolution
R. Jay Martin II
https://arxiv.org/abs/2602.11476 https://arxiv.org/pdf/2602.11476 https://arxiv.org/html/2602.11476
arXiv:2602.11476v1 Announce Type: new
Abstract: We formalize a constructive subclass of locality-preserving deterministic operators acting on graph-indexed state systems. We define the class of Bounded Local Generator Classes (BLGC), consisting of finite-range generators operating on bounded state spaces under deterministic composition. Within this class, incremental update cost is independent of total system dimension. We prove that, under the BLGC assumptions, per-step operator work satisfies W_t = O(1) as the number of nodes M \to \infty, establishing a structural decoupling between global state size and incremental computational effort. The framework admits a Hilbert-space embedding in \ell^2(V; \mathbb{R}^d) and yields bounded operator norms on admissible subspaces. The result applies specifically to the defined subclass and does not claim universality beyond the stated locality and boundedness constraints.
toXiv_bot_toot
@mgorny@social.treehouse.systemsWhile working through another last rites slew, I was thinking that back in the day there were a number of developers who believed they should add a lot of packages to #Gentoo, in the name of giving users a choice. Like, they were projects whose sole purpose of existence seemed to be to find every piece of software that roughly fit a specific topic, get it to build and package it for Gentoo.
Of course, the long-term effect of that is that there's a lot of unmaintained, often broken packages. "The choice" doesn't really work. Sure, users have a lot of packages to choose from — but they have to actually figure out which of these packages are actually useful (if any).
A few years ago attempting to remove packages also faced some verbal opposition. You shouldn't remove unmaintained or outdated packages, because they still work. You shouldn't remove packages that sometimes fail to build, because some flag combinations still work. You shouldn't remove packages that don't build at all, because the user can visit Forums and find some workaround to make them build 🤦. Or they'll have an ebuild handy to start working on it. And anyway, you shouldn't be removing stuff at all, but fixing it instead.
Sometimes the arguments were straight dishonest too: people literally said we need more packages to lure new users in. Like, it didn't matter to them that the packages didn't really work and that the people trying to use them will get a nasty surprise. They wanted people to say "hey, Gentoo has this software we need, let's start using Gentoo".
@cdarwin@c.im"We’re not just talking about potential spikes in food prices …
but also potentially key shortages in the commodities that are necessary to produce food, like fertilizers,”
says Adam Hanieh, director of the SOAS Middle East Institute at the University of London.
“Many of the countries that are going to be most potentially impacted by this are already in conditions of famine or near famine.”
@arXiv_csDS_bot@mastoxiv.pageTowards Efficient Data Structures for Approximate Search with Range Queries
Ladan Kian, Dariusz R. Kowalski
https://arxiv.org/abs/2602.06860 https://arxiv.org/pdf/2602.06860 https://arxiv.org/html/2602.06860
arXiv:2602.06860v1 Announce Type: new
Abstract: Range queries are simple and popular types of queries used in data retrieval. However, extracting exact and complete information using range queries is costly. As a remedy, some previous work proposed a faster principle, {\em approximate} search with range queries, also called single range cover (SRC) search. It can, however, produce some false positives. In this work we introduce a new SRC search structure, a $c$-DAG (Directed Acyclic Graph), which provably decreases the average number of false positives by logarithmic factor while keeping asymptotically same time and memory complexities as a classic tree structure. A $c$-DAG is a tunable augmentation of the 1D-Tree with denser overlapping branches ($c \geq 3$ children per node). We perform a competitive analysis of a $c$-DAG with respect to 1D-Tree and derive an additive constant time overhead and a multiplicative logarithmic improvement of the false positives ratio, on average. We also provide a generic framework to extend our results to empirical distributions of queries, and demonstrate its effectiveness for Gowalla dataset. Finally, we quantify and discuss security and privacy aspects of SRC search on $c$-DAG vs 1D-Tree, mainly mitigation of structural leakage, which makes $c$-DAG a good data structure candidate for deployment in privacy-preserving systems (e.g., searchable encryption) and multimedia retrieval.
toXiv_bot_toot
@arXiv_physicschemph_bot@mastoxiv.pageSpectral convergence of sum-of-Gaussians tensor neural networks for many-electron Schr\"odinger equation
Teng Wu, Qi Zhou, Huangjie Zheng, Hehu Xie, Zhenli Xu
https://arxiv.org/abs/2603.23897 https://arxiv.org/pdf/2603.23897 https://arxiv.org/html/2603.23897
arXiv:2603.23897v1 Announce Type: new
Abstract: We present an improved version of the sum-of-Gaussians tensor neural network (SOG-TNN) architecture for solving many-electron Schr\"{o}dinger equation for one-dimensional soft-Coulomb systems. Model reduction techniques are introduced to reduce the number of tensor-factorized bases under the SOG approximation of the kernel. The Slater determinant ansatz is employed so that the anti-symmetric property of the wave function can be strictly preserved. Numerical results show that the SOG-TNN achieves high accuracy with remarkably small basis sizes. Robust spectral convergence with respect to the basis size is also observed, consistently characterized by a mixed algebraic-exponential model for the error decay. These findings validate that the SOG-TNN architecture provides an ultra-efficient and low-rank representation of complex multi-electron wave functions, shedding light on high-fidelity quantum calculations in larger-scale many-electron systems.
toXiv_bot_toot
@arXiv_csCL_bot@mastoxiv.pageTraining data generation for context-dependent rubric-based short answer grading
Pavel \v{S}indel\'a\v{r}, D\'avid Slivka, Christopher Bouma, Filip Pr\'a\v{s}il, Ond\v{r}ej Bojar
https://arxiv.org/abs/2603.28537 https://arxiv.org/pdf/2603.28537 https://arxiv.org/html/2603.28537
arXiv:2603.28537v1 Announce Type: new
Abstract: Every 4 years, the PISA test is administered by the OECD to test the knowledge of teenage students worldwide and allow for comparisons of educational systems. However, having to avoid language differences and annotator bias makes the grading of student answers challenging. For these reasons, it would be interesting to compare methods of automatic student answer grading. To train some of these methods, which require machine learning, or to compute parameters or select hyperparameters for those that do not, a large amount of domain-specific data is needed. In this work, we explore a small number of methods for creating a large-scale training dataset using only a relatively small confidential dataset as a reference, leveraging a set of very simple derived text formats to preserve confidentiality. Using these methods, we successfully created three surrogate datasets that are, at the very least, superficially more similar to the reference dataset than purely the result of prompt-based generation. Early experiments suggest one of these approaches might also lead to improved model training.
toXiv_bot_toot
@arXiv_csOS_bot@mastoxiv.pageHALO: A Fine-Grained Resource Sharing Quantum Operating System
John Zhuoyang Ye, Jiyuan Wang, Yifan Qiao, Jens Palsberg
https://arxiv.org/abs/2602.07191 https://arxiv.org/pdf/2602.07191 https://arxiv.org/html/2602.07191
arXiv:2602.07191v1 Announce Type: new
Abstract: As quantum computing enters the cloud era, thousands of users must share access to a small number of quantum processors. Users need to wait minutes to days to start their jobs, which only takes a few seconds for execution. Current quantum cloud platforms employ a fair-share scheduler, as there is no way to multiplex a quantum computer among multiple programs at the same time, leaving many qubits idle and significantly under-utilizing the hardware. This imbalance between high user demand and scarce quantum resources has become a key barrier to scalable and cost-effective quantum computing.
We present HALO, the first quantum operating system design that supports fine-grained resource-sharing. HALO introduces two complementary mechanisms. First, a hardware-aware qubit-sharing algorithm that places shared helper qubits on regions of the quantum computer that minimize routing overhead and avoid cross-talk noise between different users' processes. Second, a shot-adaptive scheduler that allocates execution windows according to each job's sampling requirements, improving throughput and reducing latency. Together, these mechanisms transform the way quantum hardware is scheduled and achieve more fine-grained parallelism.
We evaluate HALO on the IBM Torino quantum computer on helper qubit intense benchmarks. Compared to state-of-the-art systems such as HyperQ, HALO improves overall hardware utilization by up to 2.44x, increasing throughput by 4.44x, and maintains fidelity loss within 33%, demonstrating the practicality of resource-sharing in quantum computing.
toXiv_bot_toot
@mgorny@social.treehouse.systems @mgorny@social.treehouse.systems
