training-resilience
Fix PPO training early-stop issues. Trigger when: (1) impossible drawdown values (>100%), (2) training stops too early, (3) need adaptive recovery instead of hard stop.
浏览 smith6jt-cop 维护的已发布 AI 技能,比较用途与说明,并通过来源仓库了解安装方式。
已收录技能
92收录技能的仓库 Stars 合计
92仓库 Forks
0平均每个技能的仓库 Stars
1同一仓库中的多个技能可能共享 Stars,合计包含重复仓库数据。
Fix PPO training early-stop issues. Trigger when: (1) impossible drawdown values (>100%), (2) training stops too early, (3) need adaptive recovery instead of hard stop.
Fix phantom MaxDD values in training by calibrating reward_scale. Trigger when: (1) validation MaxDD shows 35-80% values, (2) MaxDD doesn't correlate with training quality, (3) gating thresholds seem too lenien...
SNAKE vs RASTER tile acquisition patterns for image stitching in KINTSUGI
KINTSUGI processing principles: Never sacrifice quality for speed, always use GPU when available. Trigger: performance optimization, CPU/GPU choice, fast mode, quality vs speed.
Trading dashboard P&L visualization with profit tracker integration, win-rate overlays, R-multiples, and configurable settings
Set up conda with multiple group storage locations on HiPerGator, including cache isolation
Debugging constant Markov regime features in RL observations - when HMM probabilities show uniform values instead of dynamic regime estimates
Optimizing segmentation feature extraction, SOM clustering, and spatial analysis for multiplex immunofluorescence imaging with cell compartment and ECM features
Spatial region-aware cell matching for CODEX/scRNAseq integration
Preventing Jupyter cell hangs with explicit CuPy GPU memory cleanup
Per-sector candidate limits using yfinance sector data. Trigger when: (1) universe selection has too many candidates, (2) need sector diversity in candidate pool, (3) downloading historical data takes too long.
Testing suite with benchmarking for version-to-version performance comparison
KINTSUGI quantification plots: Always save QC heatmaps and profiles as PDF
MaxFuse parameter tuning for datasets with large protein:RNA cell ratios (>100:1)
Use when tuning predator-prey dynamics, regime detection coefficients, or cooldown mechanisms
MaxFuse parameter tuning when protein panel size changes (26 → 59+ markers)
Scale alignment for RNA-protein cross-modal integration - BOTH modalities must be z-scored
Use when tuning RL reward function weights, fixing reward hacking, or addressing P&L gradient issues
Use when implementing unified position sizing, Kelly criterion, or combining GARCH+Capital+Portfolio risk systems
Integrating SLURM job submission with kintsugi init for HPC clusters. Trigger: adding SLURM support to projects, auto-detecting HPC settings, CLI batch submission
默认按已收录数据中的 GitHub 仓库 Stars 排序。Stars 属于来源仓库,同一仓库的多个技能可能共享相同数值;它反映仓库关注度,不代表单个技能的安装量或质量评分。