--- title: Learning the Law Behind Chaos | daily.dev description: A detailed mathematical exploration of whether the governing parameter of a chaotic system (the logistic map) can be identified using the LMS adaptive... canonical: https://daily.dev/posts/learning-the-law-behind-chaos-ao6wcq5oo twitter:card: summary_large_image twitter:site: @dailydotdev og:type: website og:site_name: daily.dev og:title: Learning the Law Behind Chaos | daily.dev og:description: A detailed mathematical exploration of whether the governing parameter of a chaotic system (the logistic map) can be identified using the LMS adaptive... og:url: https://daily.dev/posts/learning-the-law-behind-chaos-ao6wcq5oo og:image: https://api.daily.dev/og/posts/Ao6wcQ5Oo.png og:image:alt: Learning the Law Behind Chaos og:image:width: 1200 og:image:height: 630 og:locale: en --- > ## Documentation Index > Fetch the complete documentation index at: https://daily.dev/llms.txt > Use this file to discover all available pages before exploring further. # Learning the Law Behind Chaos **[Medium](https://daily.dev/sources/medium_js)** · 19 min read · 0 upvotes · 0 comments ## Summary A detailed mathematical exploration of whether the governing parameter of a chaotic system (the logistic map) can be identified using the LMS adaptive filtering algorithm. Using NLMS, the true parameter r is recovered to machine precision (double-precision floating-point floor) in under 200 iterations — even at r=4 where chaos is total. Key findings: NLMS converges geometrically at rate (1−mu) per step regardless of the dynamical regime; chaos does not help and at full chaos actively worsens noise sensitivity due to the arcsine invariant density concentrating orbits near the edges where the regressor vanishes; the step size mu acts as a memory knob trading noise rejection for tracking speed; and with mu=1 and clean measurements, exact identification is achieved in a single step. ## Full article daily.dev links to this article rather than hosting it. Read it at the original source: ## Questions this post answers ### Does chaos in a dynamical system make its governing parameter easier or harder to estimate with an adaptive filter like LMS? Chaos does not help and can hurt, depending on the algorithm. Normalized LMS (NLMS) converges at nearly the same rate regardless of regime because it divides out regressor energy, but plain LMS converges slower under chaos since the regressor energy at full chaos (r=4) is less than half that at a stable fixed point, and under measurement noise the fully chaotic case becomes roughly seventy times harder to identify than a fixed point due to noise amplification near the interval edges. _Anyone weighing adaptive estimators for noisy nonlinear systems can track this kind of trade-off analysis on daily.dev._ ### Why does normalized LMS (NLMS) converge at almost the same speed regardless of whether the underlying system is at a fixed point, periodic, or fully chaotic? NLMS divides its update by the regressor energy, so the parameter error contracts by exactly (1 minus the step size mu) every step regardless of the orbit's behavior, as long as the regressor is nonzero. This cancels the orbit's statistics from the learning rate, so iterations to reach a given accuracy vary by only a couple of steps across fixed-point, periodic, and chaotic regimes. _Engineers comparing adaptive filter variants for real systems can find this kind of derivation-backed comparison on daily.dev._ ### What step size mu makes an LMS-style adaptive filter converge in a single step (deadbeat)? A step size of mu equal to 1 produces deadbeat convergence for NLMS in the noise-free case, annihilating the parameter error in exactly one observation. Stability holds for 0 less than mu less than 2, with convergence rate given by the absolute value of (1 minus mu); however, at mu=1 with any measurement noise, the estimate swallows each noisy sample whole and inherits the noise undiluted, so smaller mu trades speed for noise tolerance. _Developers tuning adaptive filter step sizes for real-world noisy signals can dig into this trade-off on daily.dev._ ## Similar posts on daily.dev - [I Built Adaptive Noise Cancellation in MATLAB Then Realized It Was Online Regression](https://daily.dev/posts/i-built-adaptive-noise-cancellation-in-matlab-then-realized-it-was-online-regression-xzqzhhdpa) · Medium · 0 upvotes · 0 comments - [How LLMs Can Be Assisted To Do Arithmetic Correctly](https://daily.dev/posts/how-llms-can-be-assisted-to-do-arithmetic-correctly-04arkfsu0) · Hackaday · 1 upvotes · 0 comments - [5 LLM Fine-tuning Techniques](https://daily.dev/posts/5-llm-fine-tuning-techniques-olu0mzhzh) · Daily Dose of Data Science \| Avi Chawla \| Substack · 3 upvotes · 0 comments - [\[2602.04118\] Learning to Reason in 13 Parameters](https://daily.dev/posts/2602-04118-learning-to-reason-in-13-parameters-fgsqvuufg) · Hacker News · 1 upvotes · 0 comments --- Tags: [#machine-learning](https://daily.dev/tags/machine-learning), [#signal-processing](https://daily.dev/tags/signal-processing) [View this post on daily.dev](https://daily.dev/posts/learning-the-law-behind-chaos-ao6wcq5oo) ```json {"@context":"https://schema.org","@graph":[{"@type":"Organization","@id":"https://daily.dev/#organization","name":"daily.dev","url":"https://daily.dev","logo":{"@type":"ImageObject","url":"https://daily.dev/apple-touch-icon.png","width":180,"height":180},"sameAs":["https://twitter.com/dailydotdev","https://github.com/dailydotdev","https://www.linkedin.com/company/daily-dev-ltd"]},{"@type":"WebSite","@id":"https://daily.dev/#website","url":"https://daily.dev","name":"daily.dev","publisher":{"@id":"https://daily.dev/#organization"},"potentialAction":{"@type":"SearchAction","target":{"@type":"EntryPoint","urlTemplate":"https://daily.dev/search?q={search_term_string}"},"query-input":"required name=search_term_string"}}]} {"@context":"https://schema.org","@type":"TechArticle","headline":"Learning the Law Behind Chaos","url":"https://daily.dev/posts/learning-the-law-behind-chaos-ao6wcq5oo","mainEntityOfPage":{"@type":"WebPage","@id":"https://daily.dev/posts/learning-the-law-behind-chaos-ao6wcq5oo"},"datePublished":"2026-07-21T16:17:39.943Z","dateModified":"2026-09-14T08:20:18.314Z","description":"A detailed mathematical exploration of whether the governing parameter of a chaotic system (the logistic map) can be identified using the LMS adaptive...","image":"https://media.daily.dev/image/upload/s--foaA6JGU--/f_auto/v1722860399/public/Placeholder%2004","thumbnailUrl":"https://media.daily.dev/image/upload/s--foaA6JGU--/f_auto/v1722860399/public/Placeholder%2004","isAccessibleForFree":true,"articleSection":"Medium","inLanguage":"en","publisher":{"@type":"Organization","name":"daily.dev","url":"https://daily.dev","logo":{"@type":"ImageObject","url":"https://daily.dev/apple-touch-icon.png","width":180,"height":180}},"author":{"@type":"Organization","name":"Medium","logo":"https://media.daily.dev/image/upload/t_logo,f_auto/v1/logos/medium","url":"https://daily.dev/sources/medium_js"},"commentCount":0,"discussionUrl":"https://daily.dev/posts/learning-the-law-behind-chaos-ao6wcq5oo","interactionStatistic":[{"@type":"InteractionCounter","interactionType":{"@type":"LikeAction"},"userInteractionCount":0},{"@type":"InteractionCounter","interactionType":{"@type":"CommentAction"},"userInteractionCount":0}],"keywords":"machine-learning,signal-processing","timeRequired":"PT19M"} {"@context":"https://schema.org","@type":"BreadcrumbList","itemListElement":[{"@type":"ListItem","position":1,"name":"Home","item":"https://daily.dev"},{"@type":"ListItem","position":2,"name":"Medium","item":"https://daily.dev/sources/medium_js"},{"@type":"ListItem","position":3,"name":"Learning the Law Behind Chaos"}]} {"@context":"https://schema.org","@type":"FAQPage","@id":"https://daily.dev/posts/learning-the-law-behind-chaos-ao6wcq5oo#faq","mainEntity":[{"@type":"Question","name":"Does chaos in a dynamical system make its governing parameter easier or harder to estimate with an adaptive filter like LMS?","acceptedAnswer":{"@type":"Answer","text":"Chaos does not help and can hurt, depending on the algorithm. Normalized LMS (NLMS) converges at nearly the same rate regardless of regime because it divides out regressor energy, but plain LMS converges slower under chaos since the regressor energy at full chaos (r=4) is less than half that at a stable fixed point, and under measurement noise the fully chaotic case becomes roughly seventy times harder to identify than a fixed point due to noise amplification near the interval edges. Anyone weighing adaptive estimators for noisy nonlinear systems can track this kind of trade-off analysis on daily.dev."}},{"@type":"Question","name":"Why does normalized LMS (NLMS) converge at almost the same speed regardless of whether the underlying system is at a fixed point, periodic, or fully chaotic?","acceptedAnswer":{"@type":"Answer","text":"NLMS divides its update by the regressor energy, so the parameter error contracts by exactly (1 minus the step size mu) every step regardless of the orbit's behavior, as long as the regressor is nonzero. This cancels the orbit's statistics from the learning rate, so iterations to reach a given accuracy vary by only a couple of steps across fixed-point, periodic, and chaotic regimes. Engineers comparing adaptive filter variants for real systems can find this kind of derivation-backed comparison on daily.dev."}},{"@type":"Question","name":"What step size mu makes an LMS-style adaptive filter converge in a single step (deadbeat)?","acceptedAnswer":{"@type":"Answer","text":"A step size of mu equal to 1 produces deadbeat convergence for NLMS in the noise-free case, annihilating the parameter error in exactly one observation. Stability holds for 0 less than mu less than 2, with convergence rate given by the absolute value of (1 minus mu); however, at mu=1 with any measurement noise, the estimate swallows each noisy sample whole and inherits the noise undiluted, so smaller mu trades speed for noise tolerance. Developers tuning adaptive filter step sizes for real-world noisy signals can dig into this trade-off on daily.dev."}}]} ```