How manh20060961815271-arch/ma-tr-n-d-b-o works
This repository is a cyclone-track forecasting experiment: it reads IBTrACS storm records, trains a neural model to predict future coordinate displacement, evaluates predictions, and produces error summaries and track maps. The main script orchestrates the workflow; training and testing share window preparation and the custom layer implementation. The shown code supports a local checkpoint store. Other saved arrays in the tree are not connected to the shown workflow. Evidence is limited to the README and sampled source excerpts.
Main entry points: MNN_Main.py Workflow orchestration · Training.py Model training · Testing.py Forecast testing · Data_Processing.py Window preparation
Components 12
Forecast workflow
- Workflow orchestration
MNN_Main.py - Storm ingestion
MNN_Main.py - Model training
Training.py - Forecast testing
Testing.py - Error evaluation
MNN_Main.py - Track map output
MNN_Main.py
Model runtime
- Window preparation
Data_Processing.py - M-neuron layer
Add_Layer.py
Data and artifacts
- IBTrACS input
ibtracs-tbd.csv - Model checkpoint
- Saved prediction arrays
Y_pred_trueMNN.npy
Other components
- Research user
Connections 13
- Research user → Workflow orchestration: runs
- Workflow orchestration → Storm ingestion: loads storms
MNN_Main.py - IBTrACS input → Storm ingestion: provides records
MNN_Main.py - Workflow orchestration → Model training: trains model
MNN_Main.py - Model training → Window preparation: prepares windows
Training.py - Model training → M-neuron layer: builds layers
Training.py - Model training → Model checkpoint: saves best model
Training.py - Workflow orchestration → Forecast testing: tests forecast
MNN_Main.py - Forecast testing → Window preparation: prepares windows
Testing.py - Forecast testing → M-neuron layer: builds layers
Testing.py - Forecast testing → Model checkpoint: restores model
Testing.py - Workflow orchestration → Error evaluation: computes errors
MNN_Main.py - Workflow orchestration → Track map output: plots tracks
MNN_Main.py