6.0.1

✨ New Features

• Agents: Added support for Anthropic, Gemini, and 3rd party models.
• Internationalization: Added support for Chinese (CN), Japanese (JP), French (FR), and Spanish (ES).
• Deep Learning Enhancements: Added new activation functions: Mish, ELU, and Hardswish.
• Data Handling: Support float[][] input data for K-Means, G-Means, and X-Means.

🔧 Improvements

• Performance & Stability:
  • Significant reduction in memory leaks and tightened memory management.
  • Improved input validation and guards across various modules (Anomaly Detection, Manifold, etc.).
  • Corrected Recursive Least Squares (RLS) forgetting-factor logic.
• API Refinement:
  • Upgraded to Vega-Lite v6 for better visualizations.
  • Mixed-precision distance() calculations in the Vector class.
• Documentation: Extensive updates to documentation including new guides for:
  • Feature Engineering, Regression, Classification, Clustering, Anomaly Detection, and Hyper-parameter Optimization.
  • Time Series and Sequence modeling.

🐛 Bug Fixes

• Core Logic:
  • Fixed Huber loss calculation.
  • Resolved index and Time Series Split (TSS) bugs.
  • Fixed InputLayer.readObject() serialization issue.
• General: Fixed various edge cases in var/stdev calculations for short arrays.

🛠 Internal & CI/CD

• CI/CD Pipeline: Added new GitHub Action workflows (ci.yml and cd.yml) for automated testing and deployment.
• Test Consistency: Standardized commonPool size for more reliable test results across different machines.