How anthimevalin/multi-trajectory-pinns-zero-terminal-hjb works
This repository is a research implementation for solving and evaluating zero-terminal HJB control problems with PINN and MT-PINN models. The principal workflow is to train or select a benchmark model, restore its configuration and checkpoint, generate learned and analytic control/value trajectories, then compare and plot results. Order Execution also supports market-data window construction and real-market policy diagnostics. The README documents these capabilities; some notebook internals and the market-data call wiring were not sampled, so those links are not inferred.
Main entry points: evaluation.ipynb LQR Evaluation · general_evaluation.ipynb Execution Evaluation · Models LQR Training · market_eval_demo.ipynb Market Demo
Components 17
High-Dimensional LQR
- LQR Training
High-Dimensional-LQR/Models - LQR Evaluation
High-Dimensional-LQR/evaluation.ipynb - LQR Analytics
High-Dimensional-LQR/compute_helper.py - LQR Plots
High-Dimensional-LQR/plot_helper.py
Order Execution
- Execution Training
Order-Execution/Models - Execution Evaluation
Order-Execution/general_evaluation.ipynb - Execution Analytics
Order-Execution/compute_helper.py - Market Demo
Order-Execution/market_eval_demo.ipynb - Market Tools
Order-Execution/market_eval.py
Models And Artifacts
- LQR Model Loader
High-Dimensional-LQR/loading_helper.py - Execution Model Loader
Order-Execution/loading_helper.py - LQR Checkpoints
High-Dimensional-LQR/runs - Execution Checkpoints
Order-Execution/runs
Evaluation And Diagnostics
- LQR Comparisons
High-Dimensional-LQR/plot_helper.py - Execution Comparisons
Order-Execution/plot_helper.py
Other components
- Researcher
- BBO Market Data
Connections 20
- Researcher → LQR Training: runs notebooks
- Researcher → LQR Evaluation: evaluates
- LQR Training → LQR Checkpoints: saves artifacts
- LQR Evaluation → LQR Model Loader: restores model
- LQR Model Loader → LQR Checkpoints: reads checkpoint
- LQR Evaluation → LQR Analytics: evaluates trajectories
- LQR Evaluation → LQR Plots: plots comparisons
- LQR Training → LQR Analytics: uses control helpers
- Researcher → Execution Training: runs notebooks
- Researcher → Execution Evaluation: evaluates
- Execution Training → Execution Checkpoints: saves artifacts
- Execution Evaluation → Execution Model Loader: restores models
- Execution Model Loader → Execution Checkpoints: reads checkpoint
- Execution Evaluation → Execution Analytics: generates trajectories
- Execution Evaluation → Execution Comparisons: plots comparisons
- Execution Analytics → Execution Comparisons: supplies trajectories
- Researcher → Market Demo: configures analysis
- Market Demo → Market Tools: uses market tools
- Market Demo → Execution Model Loader: loads policy
- BBO Market Data → Market Tools: parses quotes