{"slug": "thread-differentiable-self-organizing-systems", "title": "Thread: Differentiable Self-organizing Systems", "summary": "Distill thread exploring differentiable self-organizing systems, which use optimization to learn individual agent behaviors that achieve collective goals. It presents several research articles on topics like morphogenesis, digit self-classification, texture synthesis, and adversarial reprogramming, all using Neural Cellular Automata models. The thread is designed as a living document to facilitate cross-disciplinary exchange between machine learning and developmental biology.", "body_md": "Self-organisation is omnipresent on all scales of biological life. From complex interactions between molecules\nforming structures such as proteins, to cell colonies achieving global goals like exploration by means of the\nindividual cells collaborating and communicating, to humans forming collectives in society such as tribes,\ngovernments or countries. The old adage “the whole is greater than the sum of its parts”, often ascribed to\nAristotle, rings true everywhere we look.\nThe articles in this thread focus on practical ways of designing self-organizing systems. In particular we use\nDifferentiable Programming (optimization) to learn agent-level policies that satisfy system-level objectives. The\ncross-disciplinary nature of this thread aims to facilitate ideas exchange between ML and developmental biology\ncommunities.\nArticles & Comments\nDistill has invited several researchers to publish a “thread” of short articles exploring differentiable\nself-organizing systems,\ninterspersed with critical commentary from several experts in adjacent fields.\nThe thread will be a living document, with new articles added over time.\nArticles and comments are presented below in chronological order:\nBuilding their own bodies is the very first skill all living creatures possess. How can we design systems that\ngrow, maintain and repair themselves by regenerating damages? This work investigates morphogenesis, the\nprocess by which living creatures self-assemble their bodies. It proposes a differentiable, Cellular Automata\nmodel of morphogenesis and shows how such a model learns a robust and persistent set of dynamics to grow any\narbitrary structure starting from a single cell.\nRead Full Article\nThis work presents a follow up to Growing Neural CAs, using a similar computational model for the goal of\ndigit “self-classification”. The authors show how neural CAs can self-classify the MNIST digit they form. The\nresulting CAs can be interacted with by dynamically changing the underlying digit. The CAs respond to\nperturbations with a learned self-correcting classification behaviour.\nRead Full Article\nHere the authors apply Neural Cellular Automata to a new domain: texture synthesis. They begin by training NCA\nto mimic a series of textures taken from template images. Then, taking inspiration from adversarial\ncamouflages which appear in nature, they use NCA to create textures which maximally excite neurons in a\npretrained vision model. These results reveal that a simple model combined with well-known objectives can lead\nto robust and unexpected behaviors.\nRead Full Article\nThis work takes existing Neural CA models and shows how they can be adversarially reprogrammed to perform novel tasks.\nMNIST CA can be deceived into outputting incorrect classifications and the patterns in Growing CA can be made to have their shape and colour altered.\nRead Full Article\nThis is a living document\nExpect more articles on this topic, along with critical comments from\nexperts.\nGet Involved\nThe Self-Organizing systems thread is open to articles exploring differentiable self-organizing sytems.\nCritical\ncommentary and discussion of existing articles is also welcome. The thread\nis organized through the open #selforg\nchannel on the\nDistill slack. Articles can be\nsuggested there, and will be included at the discretion of previous\nauthors in the thread, or in the case of disagreement by an uninvolved\neditor.\nIf you would like get involved but don’t know where to start, small\nprojects may be available if you ask in the channel.\nAbout the Thread Format\nPart of Distill’s mandate is to experiment with new forms of scientific\npublishing. We believe that that reconciling faster and more continuous\napproaches to publication with review and discussion is an important open\nproblem in scientific publishing.\nThreads are collections of short articles, experiments, and critical\ncommentary around a narrow or unusual research topic, along with a slack\nchannel for real time discussion and collaboration. They are intended to\nbe earlier stage than a full Distill paper, and allow for more fluid\npublishing, feedback and discussion. We also hope they’ll allow for wider\nparticipation. Think of a cross between a Twitter thread, an academic\nworkshop, and a book of collected essays.\nThreads are very much an experiment. We think it’s possible they’re a\ngreat format, and also possible they’re terrible. We plan to trial two\nsuch threads and then re-evaluate our thought on the format.\nEditorial Note\nPart of Distill’s mandate is to experiment with new forms of scientific publishing.\nWe believe something along the lines of this “thread” format might be promising,\nbut see it very much as an experiment.\nWe plan to trial two such threads and then re-evaluate our thought on the format.\nCitation Information\nIf you wish to cite this thread as a whole, citation information can be found below.\nThe author order is all participants in the thread in alphabetical order.\nSince this is a living document, the citation may add additional authors as it evolves.\nYou can also cite individual articles using the citation information provided at the\nbottom of the corresponding article.\nUpdates and Corrections\nIf you see mistakes or want to suggest changes, please create an issue on GitHub.\nReuse\nDiagrams and text are licensed under Creative Commons Attribution CC-BY 4.0 with the source available on GitHub, unless noted otherwise. The figures that have been reused from other sources don’t fall under this license and can be recognized by a note in their caption: “Figure from …”.\nCitation\nFor attribution in academic contexts, please cite this work as\nMordvintsev, et al., \"Thread: Differentiable Self-organizing Systems\", Distill, 2020.\nBibTeX citation\n@article{mordvintsev2020thread:,\nauthor = {Mordvintsev, Alexander and Randazzo, Ettore and Niklasson, Eyvind and Levin, Michael and Greydanus, Sam},\ntitle = {Thread: Differentiable Self-organizing Systems},\njournal = {Distill},\nyear = {2020},\nnote = {https://distill.pub/2020/selforg},\ndoi = {10.23915/distill.00027}\n}", "url": "https://wpnews.pro/news/thread-differentiable-self-organizing-systems", "canonical_source": "https://distill.pub/2020/selforg", "published_at": "2020-08-27 20:00:00+00:00", "updated_at": "2026-05-19 23:17:56.364345+00:00", "lang": "en", "topics": ["artificial-intelligence", "machine-learning", "research", "science", "robotics"], "entities": ["Aristotle", "Distill"], "alternates": {"html": "https://wpnews.pro/news/thread-differentiable-self-organizing-systems", "markdown": "https://wpnews.pro/news/thread-differentiable-self-organizing-systems.md", "text": "https://wpnews.pro/news/thread-differentiable-self-organizing-systems.txt", "jsonld": "https://wpnews.pro/news/thread-differentiable-self-organizing-systems.jsonld"}}