[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"skill:01M7EV2W8WJ2HBZ1KDNTWCSG43":3,"skill-same:01M7EV2W8WJ2HBZ1KDNTWCSG43":47},{"skillNo":4,"name":5,"nameZh":6,"description":7,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":16,"categories":17,"keywords":18,"publishedAt":19,"linkOnly":20,"markdownContent":21,"markdownBody":22,"frontMatter":23,"refName":25,"commitSha":26,"skillDirectory":27,"forksCount":28,"repositoryHtmlUrl":29,"repositoryHomepageUrl":6,"ownerHtmlUrl":30,"ownerBlogUrl":31,"ownerDescription":6,"updatedAt":32,"install":33,"resources":46},"01M7EV2W8WJ2HBZ1KDNTWCSG43","neural-decoding",null,"Decoding mental states and stimuli from neural signals — classifiers, cross-validation, and leakage-free evaluation.","TanNhatCMS\u002Fjarvis","TanNhatCMS","https:\u002F\u002Favatars.githubusercontent.com\u002Fu\u002F113796420?v=4","01M753MV17KG5CD3152G3WQEQY","KietNT","01M7EKEDYVYQ6PWVJJMZ911VMX",0,"MIT","https:\u002F\u002Fgithub.com\u002FTanNhatCMS\u002Fjarvis\u002Fblob\u002Fmain\u002Fskills\u002Fams-neural-decoding\u002FSKILL.md",[],[],"2026-11-15 14:19:17",false,"---\nname: neural-decoding\ndescription: Decoding mental states and stimuli from neural signals — classifiers, cross-validation, and leakage-free evaluation.\ncategory: scientific\n---\n\n## Overview\n\nneural-decoding is the practice of predicting stimuli, decisions, or behavior from measured neural\nactivity: which image was seen, which choice will be made, where attention is directed. It spans\nfMRI MVPA, EEG\u002FMEG classification, and spike-train decoding, and it is one of the most\nleakage-prone analyses in neuroscience — most decoding \"successes\" that don't replicate trace back\nto evaluation errors, not brain signals.\n\nThis skill is about doing decoding honestly: proper cross-validation, appropriate baselines, and\ninterpretations that don't outrun the evidence.\n\n## When to use\n\n- MVPA on fMRI: searchlight or ROI-based classification of conditions.\n- EEG\u002FMEG decoding: time-resolved classification (temporal generalization matrices).\n- Spike-train decoding: population vector, Bayesian, or ML decoders.\n- Choosing classifiers: linear SVM, logistic regression, LDA — and when not to use deep nets.\n- Evaluating: cross-validation schemes, permutation-based significance, confusion matrices.\n- Interpreting weights: why raw classifier weights mislead and what to use instead.\n\n## Core concepts\n\n- **Linear first.** Linear SVM, logistic regression, and LDA dominate neural decoding because\n  neural datasets are small (tens-hundreds of trials) relative to features (thousands of voxels).\n  Deep networks rarely help and usually overfit; start linear and only add complexity with\n  evidence.\n- **Cross-validation done right.** Split by independent units: runs\u002Fsessions for fMRI (never\n  adjacent trials — temporal autocorrelation leaks), trials blocked by time for EEG, never shuffle\n  within a trial. Leave-one-run-out is the fMRI default. Report the scheme explicitly.\n- **The leakage catalog.** Feature selection on all data then CV; normalizing using all trials;\n  hyperparameter tuning on the test set; trial-shuffled splits with autocorrelated data; using\n  future time points to decode past ones. Each inflates accuracy; check for all of them.\n- **Chance is not 50%.** With imbalanced classes, permutation testing gives the true null\n  distribution — never compare to a theoretical chance level when classes are unbalanced or\n  cross-validation is non-standard. Permute labels within the exact CV scheme, ≥1000 permutations.\n- **Temporal generalization (EEG\u002FMEG).** Train at time t, test at all times t′ — the resulting\n  matrix reveals whether representations are stable (diagonal generalization) or dynamic\n  (off-diagonal structure). This is more informative than single-time-point accuracy.\n- **Searchlight.** A classifier in every local neighborhood, mapped back to the brain. Powerful\n  but expensive and smooth-looking; correct for multiple comparisons (permutation + TFCE) and\n  remember the map shows where information is, with searchlight-sized spatial uncertainty.\n- **Weight interpretation.** Raw classifier weights are not importance maps — a large weight can\n  reflect noise suppression, not signal. Use Haufe's activation-pattern transform (weights ×\n  data covariance) for interpretable maps.\n- **What decoding proves.** Above-chance decoding shows information about the label is present in\n  the measured signal. It does not show the brain uses that information, that the representation\n  is explicit, or (for fMRI) anything about timing. State exactly this.\n\n## Practical workflow\n\n1. **Define the question.** What label, what features, what would above-chance mean? Preregister\n   the classifier and CV scheme.\n2. **Prepare features.** fMRI: trial-wise betas or t-maps (not raw volumes); EEG: time-resolved\n   voltages or time-frequency power; spikes: binned counts. Normalize within training folds only.\n3. **Choose CV.** Leave-one-run-out (fMRI), blocked K-fold (EEG), stratified when classes are\n   imbalanced. No peeking: all preprocessing parameters from training data.\n4. **Baseline model.** Linear SVM or logistic regression with default-ish regularization. Tune C\n   in nested CV if you must.\n5. **Significance.** Permutation test within the same CV scheme; correct across searchlights\u002Ftime\n   points.\n6. **Interpret.** Confusion matrices (which classes confuse?), Haufe-transformed patterns,\n   temporal generalization. Report accuracy with CIs, not just \"significant.\"\n\nExample (Python sketch):\n```python\nfrom sklearn.svm import SVC\nfrom sklearn.model_selection import LeaveOneGroupOut, permutation_test_score\nclf = SVC(kernel=\"linear\")\nscore, perm, p = permutation_test_score(clf, X, y, groups=runs,\n                                        cv=LeaveOneGroupOut(), n_permutations=1000)\n```\n\n## Common pitfalls\n\n- Feature selection or normalization before the CV split (leakage).\n- Shuffled trial-wise CV on autocorrelated data.\n- Comparing to 50% chance with imbalanced classes instead of permuting.\n- Interpreting raw SVM weights as brain maps.\n- Deep nets on 200 trials presented as methodological advance.\n- Claiming the brain \"represents\" the label — decoding shows information presence, not use.\n- Searchlight maps without multiple-comparisons correction.\n","## Overview\n\nneural-decoding is the practice of predicting stimuli, decisions, or behavior from measured neural\nactivity: which image was seen, which choice will be made, where attention is directed. It spans\nfMRI MVPA, EEG\u002FMEG classification, and spike-train decoding, and it is one of the most\nleakage-prone analyses in neuroscience — most decoding \"successes\" that don't replicate trace back\nto evaluation errors, not brain signals.\n\nThis skill is about doing decoding honestly: proper cross-validation, appropriate baselines, and\ninterpretations that don't outrun the evidence.\n\n## When to use\n\n- MVPA on fMRI: searchlight or ROI-based classification of conditions.\n- EEG\u002FMEG decoding: time-resolved classification (temporal generalization matrices).\n- Spike-train decoding: population vector, Bayesian, or ML decoders.\n- Choosing classifiers: linear SVM, logistic regression, LDA — and when not to use deep nets.\n- Evaluating: cross-validation schemes, permutation-based significance, confusion matrices.\n- Interpreting weights: why raw classifier weights mislead and what to use instead.\n\n## Core concepts\n\n- **Linear first.** Linear SVM, logistic regression, and LDA dominate neural decoding because\n  neural datasets are small (tens-hundreds of trials) relative to features (thousands of voxels).\n  Deep networks rarely help and usually overfit; start linear and only add complexity with\n  evidence.\n- **Cross-validation done right.** Split by independent units: runs\u002Fsessions for fMRI (never\n  adjacent trials — temporal autocorrelation leaks), trials blocked by time for EEG, never shuffle\n  within a trial. Leave-one-run-out is the fMRI default. Report the scheme explicitly.\n- **The leakage catalog.** Feature selection on all data then CV; normalizing using all trials;\n  hyperparameter tuning on the test set; trial-shuffled splits with autocorrelated data; using\n  future time points to decode past ones. Each inflates accuracy; check for all of them.\n- **Chance is not 50%.** With imbalanced classes, permutation testing gives the true null\n  distribution — never compare to a theoretical chance level when classes are unbalanced or\n  cross-validation is non-standard. Permute labels within the exact CV scheme, ≥1000 permutations.\n- **Temporal generalization (EEG\u002FMEG).** Train at time t, test at all times t′ — the resulting\n  matrix reveals whether representations are stable (diagonal generalization) or dynamic\n  (off-diagonal structure). This is more informative than single-time-point accuracy.\n- **Searchlight.** A classifier in every local neighborhood, mapped back to the brain. Powerful\n  but expensive and smooth-looking; correct for multiple comparisons (permutation + TFCE) and\n  remember the map shows where information is, with searchlight-sized spatial uncertainty.\n- **Weight interpretation.** Raw classifier weights are not importance maps — a large weight can\n  reflect noise suppression, not signal. Use Haufe's activation-pattern transform (weights ×\n  data covariance) for interpretable maps.\n- **What decoding proves.** Above-chance decoding shows information about the label is present in\n  the measured signal. It does not show the brain uses that information, that the representation\n  is explicit, or (for fMRI) anything about timing. State exactly this.\n\n## Practical workflow\n\n1. **Define the question.** What label, what features, what would above-chance mean? Preregister\n   the classifier and CV scheme.\n2. **Prepare features.** fMRI: trial-wise betas or t-maps (not raw volumes); EEG: time-resolved\n   voltages or time-frequency power; spikes: binned counts. Normalize within training folds only.\n3. **Choose CV.** Leave-one-run-out (fMRI), blocked K-fold (EEG), stratified when classes are\n   imbalanced. No peeking: all preprocessing parameters from training data.\n4. **Baseline model.** Linear SVM or logistic regression with default-ish regularization. Tune C\n   in nested CV if you must.\n5. **Significance.** Permutation test within the same CV scheme; correct across searchlights\u002Ftime\n   points.\n6. **Interpret.** Confusion matrices (which classes confuse?), Haufe-transformed patterns,\n   temporal generalization. Report accuracy with CIs, not just \"significant.\"\n\nExample (Python sketch):\n```python\nfrom sklearn.svm import SVC\nfrom sklearn.model_selection import LeaveOneGroupOut, permutation_test_score\nclf = SVC(kernel=\"linear\")\nscore, perm, p = permutation_test_score(clf, X, y, groups=runs,\n                                        cv=LeaveOneGroupOut(), n_permutations=1000)\n```\n\n## Common pitfalls\n\n- Feature selection or normalization before the CV split (leakage).\n- Shuffled trial-wise CV on autocorrelated data.\n- Comparing to 50% chance with imbalanced classes instead of permuting.\n- Interpreting raw SVM weights as brain maps.\n- Deep nets on 200 trials presented as methodological advance.\n- Claiming the brain \"represents\" the label — decoding shows information presence, not use.\n- Searchlight maps without multiple-comparisons correction.\n",{"name":5,"category":24,"description":7},"scientific","main","af388a77c2fe5e0b4fd879cca4300a0ffa9a6579","skills\u002Fams-neural-decoding",1,"https:\u002F\u002Fgithub.com\u002FTanNhatCMS\u002Fjarvis","https:\u002F\u002Fgithub.com\u002FTanNhatCMS","https:\u002F\u002Fmrkiet.name.vn","2026-11-07 14:23:08",{"source":8,"skillName":5,"commands":34,"prompt":44,"zipAvailable":45},[35,38,41],{"tool":36,"command":37},"npx","npx skills add TanNhatCMS\u002Fjarvis --skill neural-decoding",{"tool":39,"command":40},"bunx","bunx skills add TanNhatCMS\u002Fjarvis --skill neural-decoding",{"tool":42,"command":43},"pnpm","pnpm dlx skills add TanNhatCMS\u002Fjarvis --skill neural-decoding","安装技能，技能地址：https:\u002F\u002Fgithub.com\u002FTanNhatCMS\u002Fjarvis\u002Fblob\u002Fmain\u002Fskills\u002Fams-neural-decoding\u002FSKILL.md",true,[],[48,55,63,70,77,84],{"skillNo":49,"name":50,"nameZh":6,"description":51,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":52,"categories":53,"keywords":54,"publishedAt":19},"01M7EWYJR0J8DARY7E59Z6Y9MV","web-scrape","Turn web pages into LLM-ready markdown via the Firecrawl API (renders JS, anti-bot handling). Use when browser_open\u002Ffetch fails on JS-heavy pages, or when you need clean markdown\u002Fstructured extraction from a site. Requires FIRECRAWL_API_KEY.","https:\u002F\u002Fgithub.com\u002FTanNhatCMS\u002Fjarvis\u002Fblob\u002Fmain\u002Fskills\u002Fweb-scrape\u002FSKILL.md",[],[],{"skillNo":56,"name":57,"nameZh":6,"description":58,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":59,"sourceHtmlUrl":60,"categories":61,"keywords":62,"publishedAt":19},"01M7EWYBN5P6F2AJG3FR24T9BP","design-taste-frontend","tasteskill v2, the anti-slop frontend skill. Art direction for websites, distilled from a study of 247 curated gallery sites (One Page Love, Awwwards Sites of the Day). Deals researched directions (idea, composition, type, palette, photographs, section blueprints) from a library, enforces a strict no-slop house style with a real-browser inspector, and runs a look-and-fix loop on screenshots. Use this whenever you build, design, redesign, restyle, polish or review a landing page, marketing site, portfolio, product page, one-pager, hero or any public-facing web page, in any stack (HTML\u002FCSS, React, Next.js, Vue, Svelte, Astro, Tailwind) — also when the user says a page looks generic, templated, \"AI-made\", sloppy, or asks to make it look designed, premium, award-level or less like every other site. Not for dashboards or app screens.","https:\u002F\u002Fgithub.com\u002FTanNhatCMS\u002Fjarvis\u002Fblob\u002Fmain\u002Fskills\u002Ftaste-design\u002FLICENSE","https:\u002F\u002Fgithub.com\u002FTanNhatCMS\u002Fjarvis\u002Fblob\u002Fmain\u002Fskills\u002Ftaste-design\u002FSKILL.md",[],[],{"skillNo":64,"name":65,"nameZh":6,"description":66,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":67,"categories":68,"keywords":69,"publishedAt":19},"01M7EWY8FQFNQ2GB7NZZWGSSDX","test-driven-development","Use when implementing any feature or bugfix, before writing implementation code","https:\u002F\u002Fgithub.com\u002FTanNhatCMS\u002Fjarvis\u002Fblob\u002Fmain\u002Fskills\u002Fsuperpowers-tdd\u002FSKILL.md",[],[],{"skillNo":71,"name":72,"nameZh":6,"description":73,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":74,"categories":75,"keywords":76,"publishedAt":19},"01M7EWY5SHBFRH1HDHVWM9K7TV","systematic-debugging","Use when encountering any bug, test failure, or unexpected behavior, before proposing fixes","https:\u002F\u002Fgithub.com\u002FTanNhatCMS\u002Fjarvis\u002Fblob\u002Fmain\u002Fskills\u002Fsuperpowers-debugging\u002FSKILL.md",[],[],{"skillNo":78,"name":79,"nameZh":6,"description":80,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":81,"categories":82,"keywords":83,"publishedAt":19},"01M7EWY3NM0Z0JAKS7HDZ24K66","requesting-code-review","Use when completing tasks, implementing major features, or before merging to verify work meets requirements","https:\u002F\u002Fgithub.com\u002FTanNhatCMS\u002Fjarvis\u002Fblob\u002Fmain\u002Fskills\u002Fsuperpowers-code-review\u002FSKILL.md",[],[],{"skillNo":85,"name":86,"nameZh":6,"description":87,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":88,"categories":89,"keywords":90,"publishedAt":19},"01M7EWY0WRMWX4M9ZFY0WKC593","brainstorming","You MUST use this before any creative work - creating features, building components, adding functionality, modifying behavior, or planning anything new, in software or out of it (a talk, a business, a renovation).","https:\u002F\u002Fgithub.com\u002FTanNhatCMS\u002Fjarvis\u002Fblob\u002Fmain\u002Fskills\u002Fsuperpowers-brainstorming\u002FSKILL.md",[],[]]