[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"skill:01M7DCDKBBA8PJA706HD1W3AVR":3,"skill-same:01M7DCDKBBA8PJA706HD1W3AVR":53},{"skillNo":4,"name":5,"nameZh":6,"description":7,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":9,"repositoryNo":12,"starsCount":13,"licenseSpdxId":14,"licenseFileUrl":6,"sourceHtmlUrl":15,"categories":16,"keywords":17,"publishedAt":18,"linkOnly":19,"markdownContent":20,"markdownBody":21,"frontMatter":22,"refName":23,"commitSha":24,"skillDirectory":25,"forksCount":26,"repositoryHtmlUrl":27,"repositoryHomepageUrl":28,"ownerHtmlUrl":29,"ownerBlogUrl":30,"ownerDescription":31,"updatedAt":32,"install":33,"resources":46},"01M7DCDKBBA8PJA706HD1W3AVR","tune-regularization",null,"Choose regularisation weights and thresholds for a PyLops inversion (epsRs in regularized_inversion, damp in cgls\u002Flsqr, eps in ista\u002Ffista, epsRL1s in splitbregman, sigma\u002Ftau in spgl1, niter as implicit regularisation) with a principled, reproducible sweep - L-curve, discrepancy principle, GCV for small problems, or a synthetic test with a known model. Use when the user asks \"how do I choose epsilon \u002F lambda \u002F damping\", \"my inversion is too smooth \u002F too noisy\", \"make an L-curve\", or after solve-inverse-problem produced a working but untuned inversion.","PyLops\u002Fpylops-agent-skills","PyLops","https:\u002F\u002Favatars.githubusercontent.com\u002Fu\u002F74298005?v=4","01M7DCC0YV3FVDPP4GR6V7F43H","01M7DCC0ZTN5K09ZH48X951TBX",1,"LGPL-2.1","https:\u002F\u002Fgithub.com\u002FPyLops\u002Fpylops-agent-skills\u002Fblob\u002Fmain\u002Fplugins\u002Fpylops\u002Fskills\u002Ftune-regularization\u002FSKILL.md",[],[],"2026-11-15 14:19:17",false,"---\nname: tune-regularization\ndescription: Choose regularisation weights and thresholds for a PyLops inversion (epsRs in regularized_inversion, damp in cgls\u002Flsqr, eps in ista\u002Ffista, epsRL1s in splitbregman, sigma\u002Ftau in spgl1, niter as implicit regularisation) with a principled, reproducible sweep - L-curve, discrepancy principle, GCV for small problems, or a synthetic test with a known model. Use when the user asks \"how do I choose epsilon \u002F lambda \u002F damping\", \"my inversion is too smooth \u002F too noisy\", \"make an L-curve\", or after solve-inverse-problem produced a working but untuned inversion.\n---\n\nGoal: a **sweep script** in the user's project that evaluates a regularisation\nparameter over a range, applies one or more selection criteria, and reports the\nchosen value with a plot that justifies it. PyLops has no built-in L-curve \u002F GCV \u002F\ndiscrepancy helper, so the skill provides `reference\u002Flcurve_template.py`.\n\nRequires a working inversion (operator passes the dot-test, solver runs). If there\nis none, use `solve-inverse-problem` first.\n\n## 0. Check the environment\n\nSame as the other PyLops skills: use the project's Python runner, check\n`pylops.__version__`, inspect solver signatures in the installed version.\n\n## 1. Identify the parameter(s)\n\nMap the user's solver to the knob being tuned and its meaning:\n\n| Solver | Parameter | Objective term |\n|---|---|---|\n| `cgls` \u002F `lsqr` | `damp` | `damp² ‖x‖²` (for `lsqr` only with `x0=None`, see step 3) |\n| `regularized_inversion` | `epsRs=[ε₁, ε₂, …]` | `Σ εᵢ² ‖Rᵢ x‖²` |\n| `normal_equations_inversion` | `epsI`, `epsRs` (and `epsNRs`) | `epsI² ‖x‖²`, `Σ εᵢ² ‖Rᵢ x‖²` (same scaling as `regularized_inversion`) |\n| `ista` \u002F `fista` | `eps` | `eps ‖SOpᴴ x‖₁` (effective threshold is `eps·alpha\u002F2`) |\n| `splitbregman` | `epsRL1s`, `epsRL2s`, `mu` | L1 and L2 weights, data weight |\n| `spgl1` | `sigma` (BPDN) or `tau` (LASSO) | residual bound or L1-ball radius |\n| any Krylov solver without explicit regularisation | `niter` | early stopping (semi-convergence) |\n\nWith more than one weight, tune one at a time (fix the others at a\nscale-based guess), or fix their ratio and sweep an overall factor. Do not\nlaunch a full grid over several parameters unless the problem is small.\n\n## 2. Pick the criterion\n\nChoose based on what the user knows:\n\n1. **Known noise level** (sigma per sample, or an estimate from a noise-only\n   window): **discrepancy principle** - choose the largest regularisation whose\n   data residual `‖y − Op x‖` is ≤ `τ · σ · sqrt(N)`, `τ ≈ 1.0–1.1`. For\n   `spgl1`, this is exactly `sigma = τ σ sqrt(N)`.\n2. **Synthetic with a true model**: maximise `pylops.utils.metrics.snr(xtrue, x)`\n   (report it, but warn it does not transfer 1:1 to field data).\n3. **Nothing known**: **L-curve** - plot `log ‖Rx‖` vs `log ‖y − Op x‖` and pick the\n   point of maximum curvature. Needs a wide enough range to show both branches.\n4. **Small problem** (`Op.todense()` affordable, N, M ≲ 3000) and quadratic\n   regularisation: **GCV** computed exactly on the dense matrices (implemented in\n   the template).\n\nReport more than one criterion when they are cheap; disagreement between them is\ninformation for the user.\n\n## 3. Write and run the sweep\n\nAdapt `reference\u002Flcurve_template.py`:\n\n- replace `build_problem` with the user's operator, data and regularisation;\n- replace `solve(eps, x0)` with the user's solver call **with everything else\n  fixed** (same `niter`, tolerance, `x0` policy);\n- use a **log-spaced grid** spanning at least 4 orders of magnitude around a\n  scale-based guess (e.g. `ε₀ ≈ ‖Opᴴ y‖ \u002F ‖Rᴴ R Opᴴ y‖` for quadratic terms, or\n  `eps₀ ≈ max|Opᴴ y|` scale for L1 thresholds, where larger values give all-zero\n  models);\n- solve from large to small ε and **warm start** from the previous solution\n  (`x0=...`) to save iterations - unless the solver ignores `x0`. **Exception:\n  `damp` of `lsqr`** (PyLops and SciPy): with `x0` it solves for the correction\n  `x - x0`, so it penalises `damp² ‖x - x0‖²`, not `damp² ‖x‖²`, and a\n  warm-started sweep is biased (residual too low, discrepancy pick too large).\n  Use `x0=None` for every `lsqr(damp=...)` solve, or put the penalty in the\n  operator (`VStack([Op, eps*Reg])`, `regularized_inversion`), where `x0` is\n  safe. `cgls(damp=...)` and `epsRs` are not affected (checked numerically);\n- keep `niter` large enough that each solve has converged, otherwise the curve\n  measures iteration count, not regularisation (check a couple of cost curves);\n- for early stopping as regulariser, run once with a callback that stores\n  `x` every k iterations and compute the same criteria on iteration number.\n\nRun it, then look at the plot before trusting the automatic pick: an L-curve with\nno clear corner (common with L1 terms or very noisy data) should be reported as\nsuch, not forced. With derivative regularisers a second knee often appears at\nvery large ε (the model collapses into the null space of `R`, e.g. a linear\ntrend for `SecondDerivative`); it bends the other way and is not the L-curve\ncorner – the template's curvature orientation already excludes it.\n\n## 4. Hand off\n\nReport:\n\n- the parameter swept, the range and the number of solves;\n- the value picked by each criterion and the one recommended, with the reason;\n- the plot file(s) (L-curve, residual vs ε, SNR vs ε if available) and the\n  model at the chosen value next to one over- and one under-regularised model;\n- the exact line to change in the user's inversion script.\n","Goal: a **sweep script** in the user's project that evaluates a regularisation\nparameter over a range, applies one or more selection criteria, and reports the\nchosen value with a plot that justifies it. PyLops has no built-in L-curve \u002F GCV \u002F\ndiscrepancy helper, so the skill provides `reference\u002Flcurve_template.py`.\n\nRequires a working inversion (operator passes the dot-test, solver runs). If there\nis none, use `solve-inverse-problem` first.\n\n## 0. Check the environment\n\nSame as the other PyLops skills: use the project's Python runner, check\n`pylops.__version__`, inspect solver signatures in the installed version.\n\n## 1. Identify the parameter(s)\n\nMap the user's solver to the knob being tuned and its meaning:\n\n| Solver | Parameter | Objective term |\n|---|---|---|\n| `cgls` \u002F `lsqr` | `damp` | `damp² ‖x‖²` (for `lsqr` only with `x0=None`, see step 3) |\n| `regularized_inversion` | `epsRs=[ε₁, ε₂, …]` | `Σ εᵢ² ‖Rᵢ x‖²` |\n| `normal_equations_inversion` | `epsI`, `epsRs` (and `epsNRs`) | `epsI² ‖x‖²`, `Σ εᵢ² ‖Rᵢ x‖²` (same scaling as `regularized_inversion`) |\n| `ista` \u002F `fista` | `eps` | `eps ‖SOpᴴ x‖₁` (effective threshold is `eps·alpha\u002F2`) |\n| `splitbregman` | `epsRL1s`, `epsRL2s`, `mu` | L1 and L2 weights, data weight |\n| `spgl1` | `sigma` (BPDN) or `tau` (LASSO) | residual bound or L1-ball radius |\n| any Krylov solver without explicit regularisation | `niter` | early stopping (semi-convergence) |\n\nWith more than one weight, tune one at a time (fix the others at a\nscale-based guess), or fix their ratio and sweep an overall factor. Do not\nlaunch a full grid over several parameters unless the problem is small.\n\n## 2. Pick the criterion\n\nChoose based on what the user knows:\n\n1. **Known noise level** (sigma per sample, or an estimate from a noise-only\n   window): **discrepancy principle** - choose the largest regularisation whose\n   data residual `‖y − Op x‖` is ≤ `τ · σ · sqrt(N)`, `τ ≈ 1.0–1.1`. For\n   `spgl1`, this is exactly `sigma = τ σ sqrt(N)`.\n2. **Synthetic with a true model**: maximise `pylops.utils.metrics.snr(xtrue, x)`\n   (report it, but warn it does not transfer 1:1 to field data).\n3. **Nothing known**: **L-curve** - plot `log ‖Rx‖` vs `log ‖y − Op x‖` and pick the\n   point of maximum curvature. Needs a wide enough range to show both branches.\n4. **Small problem** (`Op.todense()` affordable, N, M ≲ 3000) and quadratic\n   regularisation: **GCV** computed exactly on the dense matrices (implemented in\n   the template).\n\nReport more than one criterion when they are cheap; disagreement between them is\ninformation for the user.\n\n## 3. Write and run the sweep\n\nAdapt `reference\u002Flcurve_template.py`:\n\n- replace `build_problem` with the user's operator, data and regularisation;\n- replace `solve(eps, x0)` with the user's solver call **with everything else\n  fixed** (same `niter`, tolerance, `x0` policy);\n- use a **log-spaced grid** spanning at least 4 orders of magnitude around a\n  scale-based guess (e.g. `ε₀ ≈ ‖Opᴴ y‖ \u002F ‖Rᴴ R Opᴴ y‖` for quadratic terms, or\n  `eps₀ ≈ max|Opᴴ y|` scale for L1 thresholds, where larger values give all-zero\n  models);\n- solve from large to small ε and **warm start** from the previous solution\n  (`x0=...`) to save iterations - unless the solver ignores `x0`. **Exception:\n  `damp` of `lsqr`** (PyLops and SciPy): with `x0` it solves for the correction\n  `x - x0`, so it penalises `damp² ‖x - x0‖²`, not `damp² ‖x‖²`, and a\n  warm-started sweep is biased (residual too low, discrepancy pick too large).\n  Use `x0=None` for every `lsqr(damp=...)` solve, or put the penalty in the\n  operator (`VStack([Op, eps*Reg])`, `regularized_inversion`), where `x0` is\n  safe. `cgls(damp=...)` and `epsRs` are not affected (checked numerically);\n- keep `niter` large enough that each solve has converged, otherwise the curve\n  measures iteration count, not regularisation (check a couple of cost curves);\n- for early stopping as regulariser, run once with a callback that stores\n  `x` every k iterations and compute the same criteria on iteration number.\n\nRun it, then look at the plot before trusting the automatic pick: an L-curve with\nno clear corner (common with L1 terms or very noisy data) should be reported as\nsuch, not forced. With derivative regularisers a second knee often appears at\nvery large ε (the model collapses into the null space of `R`, e.g. a linear\ntrend for `SecondDerivative`); it bends the other way and is not the L-curve\ncorner – the template's curvature orientation already excludes it.\n\n## 4. Hand off\n\nReport:\n\n- the parameter swept, the range and the number of solves;\n- the value picked by each criterion and the one recommended, with the reason;\n- the plot file(s) (L-curve, residual vs ε, SNR vs ε if available) and the\n  model at the chosen value next to one over- and one under-regularised model;\n- the exact line to change in the user's inversion script.\n",{"name":5,"description":7},"main","a0aead16681bea6df74af88096782ad6ab4eddfb","plugins\u002Fpylops\u002Fskills\u002Ftune-regularization",0,"https:\u002F\u002Fgithub.com\u002FPyLops\u002Fpylops-agent-skills","https:\u002F\u002Fpylops.github.io\u002Fpylops-agent-skills\u002F","https:\u002F\u002Fgithub.com\u002FPyLops","https:\u002F\u002Fpylops.github.io","Matrix-Free linear algebra and optimization in Python","2026-11-02 05:28:21",{"source":8,"skillName":5,"commands":34,"prompt":44,"zipAvailable":45},[35,38,41],{"tool":36,"command":37},"npx","npx skills add PyLops\u002Fpylops-agent-skills --skill tune-regularization",{"tool":39,"command":40},"bunx","bunx skills add PyLops\u002Fpylops-agent-skills --skill tune-regularization",{"tool":42,"command":43},"pnpm","pnpm dlx skills add PyLops\u002Fpylops-agent-skills --skill tune-regularization","安装技能，技能地址：https:\u002F\u002Fgithub.com\u002FPyLops\u002Fpylops-agent-skills\u002Fblob\u002Fmain\u002Fplugins\u002Fpylops\u002Fskills\u002Ftune-regularization\u002FSKILL.md",true,[47],{"path":48,"resourceType":49,"resourceTypeDesc":50,"sizeBytes":51,"sourceHtmlUrl":52},"reference\u002Flcurve_template.py","SCRIPT","脚本",5935,"https:\u002F\u002Fgithub.com\u002FPyLops\u002Fpylops-agent-skills\u002Fblob\u002Fmain\u002Fplugins\u002Fpylops\u002Fskills\u002Ftune-regularization\u002Freference\u002Flcurve_template.py",[54,61,68,75,82,89],{"skillNo":55,"name":56,"nameZh":6,"description":57,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":9,"repositoryNo":12,"starsCount":13,"licenseSpdxId":14,"licenseFileUrl":6,"sourceHtmlUrl":58,"categories":59,"keywords":60,"publishedAt":18},"01M7DCE0ERZ2MCH1ANEBTC1Y40","solve-nonsmooth-problem","Use this skill whenever the user wants to solve an inverse problem on PyLops operators with a non-smooth, constrained or non-convex term - load it before writing the objective, choosing a solver or writing optimisation code, even when a SciPy bounded solver or a hand-written FISTA\u002FADMM loop would also work. Covers total variation, L1\u002Fsparsity in a transform, joint sparsity across signals sharing the same components, bounds\u002Fpositivity\u002Fball\u002Faffine constraints, low-rank matrix completion (nuclear norm) and plug-and-play priors with a denoiser (median filter, BM3D), solved with PyProximal (ProximalGradient, ADMM, PrimalDual, ...), step sizes from operator norms. E.g. \"add TV regularisation\", \"keep the model within physical bounds\", \"recover several signals that share their components\", \"fill in missing entries of a low-rank matrix\", \"use PyProximal\". Not for purely quadratic regularisation (Tikhonov, damping, derivative smoothing), which is plain least squares - use the pylops skill solve-inverse-problem.","https:\u002F\u002Fgithub.com\u002FPyLops\u002Fpylops-agent-skills\u002Fblob\u002Fmain\u002Fplugins\u002Fpyproximal\u002Fskills\u002Fsolve-nonsmooth-problem\u002FSKILL.md",[],[],{"skillNo":62,"name":63,"nameZh":6,"description":64,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":9,"repositoryNo":12,"starsCount":13,"licenseSpdxId":14,"licenseFileUrl":6,"sourceHtmlUrl":65,"categories":66,"keywords":67,"publishedAt":18},"01M7DCDT6XDJPE8W9JPVAAX50X","wrap-operator","Wrap a user's own forward model as a PyLops linear operator inside their project (not inside the PyLops library), with a verified adjoint, a dot-test and a small test suite. Use when the user wants to turn existing forward\u002Fadjoint code, a script at a URL, or a mathematical description into a PyLops operator they can pass to PyLops solvers (e.g. \"make my blurring function a PyLops operator\", \"wrap this forward model so I can invert it with lsqr\", \"I have a forward and adjoint, plug them into PyLops\").","https:\u002F\u002Fgithub.com\u002FPyLops\u002Fpylops-agent-skills\u002Fblob\u002Fmain\u002Fplugins\u002Fpylops\u002Fskills\u002Fwrap-operator\u002FSKILL.md",[],[],{"skillNo":69,"name":70,"nameZh":6,"description":71,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":9,"repositoryNo":12,"starsCount":13,"licenseSpdxId":14,"licenseFileUrl":6,"sourceHtmlUrl":72,"categories":73,"keywords":74,"publishedAt":18},"01M7DCDBNA6XGWDZQ1A15CG1S6","solve-inverse-problem","Use this skill whenever the user has (or can build) a PyLops operator and data and wants to invert it - load it before choosing a solver or writing any inversion code, even when the right solver or regularisation seems obvious. It sets up and runs the inversion in the user's project - choose the solver (lsqr, cgls, cg, regularized_inversion, preconditioned_inversion, fista, spgl1, omp, irls, splitbregman), the regularisation and the stopping parameters, and write a runnable script that reports convergence. E.g. \"invert my data with this operator\", \"which PyLops solver should I use\", \"add Tikhonov \u002F smoothness \u002F sparsity regularisation to my inversion\", \"solve y = Op x\". Not for an inversion that already misbehaves (use debug-inversion) or for choosing regularisation weights with an L-curve or sweep (use tune-regularization).","https:\u002F\u002Fgithub.com\u002FPyLops\u002Fpylops-agent-skills\u002Fblob\u002Fmain\u002Fplugins\u002Fpylops\u002Fskills\u002Fsolve-inverse-problem\u002FSKILL.md",[],[],{"skillNo":76,"name":77,"nameZh":6,"description":78,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":9,"repositoryNo":12,"starsCount":13,"licenseSpdxId":14,"licenseFileUrl":6,"sourceHtmlUrl":79,"categories":80,"keywords":81,"publishedAt":18},"01M7DCD5Q5P09A2R28QWWVSPGM","sliding-patch-processing","Set up PyLops windowed operators - Sliding1D, Sliding2D, Sliding3D, Patch2D, Patch3D - to apply a local transform or operator over overlapping windows of 1-D\u002F2-D\u002F3-D data, choosing window size, overlap, taper and model size with the *_design and *_pad_to_next helpers, checking coverage and the dot-test, and using them for local denoising, interpolation or inversion. Use when the user asks \"apply FFT\u002FRadon\u002Fany operator in overlapping windows\", \"local transform \u002F patching with PyLops\", \"Patch2D gives artefacts at the edges\", \"how do I pick nwin, nover, nop\".","https:\u002F\u002Fgithub.com\u002FPyLops\u002Fpylops-agent-skills\u002Fblob\u002Fmain\u002Fplugins\u002Fpylops\u002Fskills\u002Fsliding-patch-processing\u002FSKILL.md",[],[],{"skillNo":83,"name":84,"nameZh":6,"description":85,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":9,"repositoryNo":12,"starsCount":13,"licenseSpdxId":14,"licenseFileUrl":6,"sourceHtmlUrl":86,"categories":87,"keywords":88,"publishedAt":18},"01M7DCCXVMV97MEBGWSTETYPYJ","pick-sparsifying-transform","Choose a sparsifying transform for a signal or image in a PyLops sparsity-promoting inversion (FFT\u002FFFT2D\u002FFFTND, DCT, DWT\u002FDWT2D\u002FDWTND, DTCWT, Seislet, curvelets via UDCT, Radon\u002FFourierRadon\u002FChirpRadon, windowed versions with Sliding\u002FPatch), check its optional dependencies and frame properties, compare candidates quantitatively with a best-k-term compressibility test, and wire the winner into fista\u002Fspgl1 in synthesis or analysis form. Use when the user asks \"which transform should I use for sparsity\", \"wavelets or curvelets for my data\", \"make my model sparse for fista\u002Fspgl1\", \"compressive sensing \u002F interpolation with PyLops\".","https:\u002F\u002Fgithub.com\u002FPyLops\u002Fpylops-agent-skills\u002Fblob\u002Fmain\u002Fplugins\u002Fpylops\u002Fskills\u002Fpick-sparsifying-transform\u002FSKILL.md",[],[],{"skillNo":90,"name":91,"nameZh":6,"description":92,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":9,"repositoryNo":12,"starsCount":13,"licenseSpdxId":14,"licenseFileUrl":6,"sourceHtmlUrl":93,"categories":94,"keywords":95,"publishedAt":18},"01M7DCCNBV43MY51NMRKC9K7J3","debug-inversion","Use this skill whenever a user shares PyLops code or results that misbehave and asks what is wrong - load it first, before running the user's code, testing a hypothesis or answering, even when the bug looks obvious from the code. Covers a crash or traceback (dtype\u002Fcasting errors, shape mismatches), a solver (lsqr, cgls, cg, fista, ista, spgl1, regularized_inversion, splitbregman) that stalls, diverges, returns NaNs, zeros, ringing or a wrong model, one dataset or block being ignored in a stacked\u002Fjoint inversion, or a custom FunctionOperator\u002FLinearOperator that may have a wrong adjoint. Diagnoses with evidence (block-by-block dot-tests incl. complex, dtype and dims checks, operator\u002Fdata scaling, solver and eps\u002Fdamp choice), then proposes the smallest fix and verifies it. Not for choosing a solver for a new, working setup (use solve-inverse-problem) or picking regularisation weights (use tune-regularization).","https:\u002F\u002Fgithub.com\u002FPyLops\u002Fpylops-agent-skills\u002Fblob\u002Fmain\u002Fplugins\u002Fpylops\u002Fskills\u002Fdebug-inversion\u002FSKILL.md",[],[]]