gradient-methods — Claude Code skill for optimization
gradient-methods is a Claude Code skill that helps choose gradient-based optimization methods and step sizes, and diagnose convergence in minimization problems.
MIT ★ 3,951 updated 2026-01-26
npx skills add parcadei/Continuous-Claude-v3 --skill gradient-methods -a claude-codeWhat gradient-methods does
The skill covers gradient descent, step-size selection, accelerated methods, and Newton’s method. It compares fixed steps with backtracking and exact line search, and points to conjugate gradients for quadratic functions. BFGS appears as a quasi-Newton option.
Runnable examples call scipy.optimize.minimize with BFGS and CG; another command calculates a gradient with sympy_compute.py. The skill permits Bash and Read. The optimization examples need SciPy installed, so those commands will fail without that library.
For convergence checks, it suggests tracking gradient norm and function values, while watching for oscillation. An oversized step can cause oscillation, so step selection is a practical part of using the method, not just a formula to copy.
What it helps with
- choose a step-size rule for gradient descent
- compare BFGS with conjugate gradients
- calculate a function gradient with SymPy
- check convergence using gradients and function values
Example prompts for the agent
- Help me choose a step size for gradient descent on this optimization problem.
- Compare BFGS and conjugate gradients for my function.
- Calculate the function's gradient and show me how to check convergence.
How to install gradient-methods
1Skills CLI (Vercel)
npx skills add parcadei/Continuous-Claude-v3 --skill gradient-methods -a claude-code2Manually: put the files in place
~/.claude/skills/gradient-methods/SKILL.mdCopy the whole skill folder; for one project use .claude/skills/gradient-methods/
1Skills CLI (Vercel)
npx skills add parcadei/Continuous-Claude-v3 --skill gradient-methods -a codex2Manually: put the files in place
.agents/skills/gradient-methods/SKILL.mdFor all projects use ~/.agents/skills/
1Skills CLI (Vercel)
npx skills add parcadei/Continuous-Claude-v3 --skill gradient-methods -a cursor2Manually: put the files in place
.cursor/skills/gradient-methods/SKILL.mdCursor also reads .agents/skills/ and .claude/skills/
1GitHub CLI
gh skill install parcadei/Continuous-Claude-v3 .claude/skills/math/optimization/gradient-methods2Manually: put the files in place
.github/skills/gradient-methods/SKILL.md1Skills CLI (Vercel)
npx skills add parcadei/Continuous-Claude-v3 --skill gradient-methods -a gemini-cli2Platform CLI
gemini skills install https://github.com/parcadei/Continuous-Claude-v3 --path .claude/skills/math/optimization/gradient-methodsManually: put the files in place
.devin/skills/gradient-methods/SKILL.mdThe legacy .windsurf/skills/ path still works
Claude.ai
Zip the folder and upload it in Settings → Capabilities → Skills
Step by step, with commands for each platform: How to install skills in Claude Code: five ways
Questions about gradient-methods
How do I install gradient-methods in Claude Code?
Run npx skills add parcadei/Continuous-Claude-v3 --skill gradient-methods -a claude-code. The skill file is at ~/.claude/skills/gradient-methods/SKILL.md.
What does gradient-methods do?
It guides method selection, step-size choices, and convergence checks for optimization. Its coverage includes gradient descent, accelerated methods, and Newton’s method.
What do the gradient-methods examples require?
The skill uses Bash and Read. Its optimization commands call scipy.optimize.minimize, while the gradient example uses sympy_compute.py.
Source text
SKILL.md · MIT
---
name: gradient-methods
description: "Problem-solving strategies for gradient methods in optimization"
allowed-tools: [Bash, Read]
---
# Gradient Methods
## When to Use
Use this skill when working on gradient-methods problems in optimization.
## Decision Tree
1. **Basic Gradient Descent**
- Update: x_{k+1} = x_k - alpha * grad f(x_k)
- Step size alpha: fixed, diminishing, or line search
- Convergence: O(1/k) for convex, linear for strongly convex
2. **Step Size Selection**
| Method | Approach |
|--------|----------|
| Fixed | alpha constant (requires tuning) |
| Backtracking | Armijo condition: f(x - alpha*grad) <= f(x) - c*alpha*||grad||^2 |
| Exact line search | minimize f(x - alpha*grad) over alpha |
| Adaptive | Adam, RMSprop (ML applications) |
3. **Accelerated Methods**
- Momentum: add velocity term
- Nesterov: look-ahead gradient
- Conjugate gradient: for quadratic functions
- `scipy.optimize.minimize(f, x0, method='CG')` - conjugate gradient
4. **Newton's Method**
- Update: x_{k+1} = x_k - H^{-1} * grad f
- Requires Hessian (expensive but quadratic convergence)
- Quasi-Newton (BFGS): approximate Hessian
- `scipy.optimize.minimize(f, x0, method='BFGS')`
5. **Convergence Diagnostics**
- Monitor ||grad f|| < tolerance
- Check function value decrease
- Watch for oscillation (step size too large)
- `sympy_compute.py diff "f" --var x` for gradient
## Tool Commands
### Scipy_Bfgs
```bash
uv run python -c "from scipy.optimize import minimize; res = minimize(lambda x: (x[0]-1)**2 + 100*(x[1]-x[0]**2)**2, [0, 0], method='BFGS'); print('Rosenbrock min at', res.x)"
```
### Scipy_Cg
```bash
uv run python -c "from scipy.optimize import minimize; res = minimize(lambda x: x[0]**2 + x[1]**2, [1, 1], method='CG'); print('Min at', res.x)"
```
### Sympy_Gradient
```bash
uv run python -m runtime.harness scripts/sympy_compute.py diff "x**2 + y**2" --var "[x, y]"
```
## Key Techniques
*From indexed textbooks:*
- [nonlinear programming_tif] Gradient Methods** - These methods use gradient information to iteratively approach the optimum. Convergence** - Addressing convergence properties. Descent Directions and Stepsize Rules:** Focuses on how to choose descent directions and appropriate step sizes.
- [nonlinear programming_tif] The application of gradient methods to unconstrained optimal control prob- lems is straightforward in principle. For example the steepest descent method takes the form W = b oMV H, (kb ph,y), i=0,. Pl = Thus, given u¥, one computes zF by forward propagation of the system equation, and then p*¥ by backward propagation of the adjoint equation.
- [nonlinear programming_tif] Footer or Trailing Row**: - There is an empty concluding element indicated by a single ". Overall, this table serves as an index for chapters or sections within a document, with particular emphasis on optimization methods and related mathematical strategies, as evidenced by the listed methods like Gradient, Newton, and other derivative techniques. The scattered letters and empty slots may denote a form of stylistic or formatting choice rather than meaningful content in this context.
- [nonlinear programming_tif] Zoutendijk’s method uses tw ) oscalatse)Oand'ye 0,1), a i ! P, where ¢ — Y™k € and my is the firs onnegative k ok 28 %, ) it T #(z*,7"e) < -y (a) Show that (b) Prove that {d*} is gradient relat ishi i i Tt pones A related, thus establishing stationarity of the 2. Min-H Method for Optimal Control) Consider the problem of findin g sequences u = (z1,22,.
- [nonlinear programming_tif] Mustration of the function f of Exercise 1. Stability) (www) We are often interested in whether optimal solutions change radically when the problem data are slightly perturbed. This issue is addressed by stability analysis, to be contrasted with sensitivity analysis, which deals with how much optimal solutions change when problem data change.
## Cognitive Tools Reference
See `.claude/skills/math-mode/SKILL.md` for full tool documentation.
© parcadei · MIT · parcadei/Continuous-Claude-v3