yzj623/pplab/apt-atomic-loss

Claude Code and agent skill by yzj623/pplab

Trust tier 3
  • community
Source check pending

Checking that the SKILL.md file is still in the repository.

Evaluate equation-6 atomic APT categorical probabilities and density-ratio losses for finite parameter atom sets.

Install

npx vskill@latest install yzj623/pplab/apt-atomic-loss

SKILL.md

The same SKILL.md was published earlier as vectorspacelab/arex-skill/apt-atomic-loss.

APT Atomic Loss

Use this skill when APT cannot evaluate the transformed-density normalizer in closed form and a recovery can instead evaluate finite atom sets. The skill computes equation-6 probabilities and the cross-entropy loss for the generating parameter.

Inputs

  • A finite atom set Theta.
  • Posterior-estimator log densities log q(theta_i | x).
  • Prior log densities log p(theta_i).
  • Index of the atom that generated the observation.

Outputs

  • Normalized atom probabilities.
  • Atomic negative log likelihood.
  • Ratio diagnostics showing whether probability ratios match posterior/prior score ratios.

Workflow

  1. Validate matching array lengths, a valid true index, and finite prior density for the true atom.
  2. Compute scores log_q_i - log_prior_i.
  3. Normalize scores with log-sum-exp.
  4. Return -log(probability[true_index]).
  5. Use ratio diagnostics for mechanism checks in reduced recovery.

Validation

Run:

bash
python tests/test_atomic_loss.py

The tests verify normalization, constant-shift invariance, score ordering, loss reduction after a targeted update, and rejection of a true atom outside prior support.

Limitations

  • This helper is deterministic and framework-free. It does not backpropagate through a neural network.
  • Full APT with MAFs or MDNs should use the same probability semantics but would rely on a deep learning framework for gradients.

Security

  • Security scanPassed, 90 of 100Pattern checks on the skill's files.
  • Model reviewPassed

Quality evaluation

  • Task evaluationNot runNo evaluation yet. Run one in Skill Studio.

Source

Popularity

7-day trend score
5

Not refreshed yet.

Works with

Any agent vskill installs to: 56 targets in all. These 39 have a page on this site.

  • Claude Code
  • Cursor
  • GitHub Copilot
  • Windsurf
  • Codex
  • Gemini CLI
  • Cline
  • Amp
  • Roo Code
  • Goose
  • Continue
  • Kilo Code
  • Junie
  • Trae
  • OpenHands
  • Qwen Code
  • AI Agent
  • Tabnine Chat
  • Sourcegraph Cody
  • Amazon Q
  • Firebase Genkit
  • Sweep
  • Devin
  • Phind
  • Mentat
  • Claude Desktop
  • Aider
  • GPT Pilot
  • Plandex
  • bolt.new
  • v0
  • Replit Agent
  • Anse
  • TypingMind
  • BetterChatGPT
  • LibreChat
  • Warp AI
  • Pieces
  • Fig AI
Embed badge
vskill VERIFIED badge
[![vskill VERIFIED](https://verified-skill.com/api/v1/skills/yzj623/pplab/apt-atomic-loss/badge)](https://verified-skill.com/skills/yzj623/pplab/apt-atomic-loss)

Something wrong with this skill or its page? Tell us.