Project Layout
Here is the layout for files in this projects, it helps one to locate the wanted scripts and debug with ease.
The model backend by Tensorflow library
This category contains models that we are actively developing.
The Core
AnnEconForecast/core
This directory contains all scripts handling tasks including data-preprocessing, model building and visualization.
Note: generally one does not need to modify files under this directory the only wanted is replicating the results in the demonstration or run existing models on a different dataset (but in the same format, for more information on the data used, refer to data instruction).
Constants
AnnEconForecast/constants.py
This file holds all user-specified constants in a python dictionary fashion.
Where our datasets go
AnnEconForecast/annefdata
By default, we placed all dataset files under this directory. Placing dataset there is not a strict requirement since our model, by default settings, would load the specified dataset via an absolute path instead of a relative one.
Demonstration
AnnEconForecast/demo
This folder holds source codes, visualizations and saved model files for demo purposes.
Configurations for HPT
AnnEconForecast/hpt_configs
Configurations for hyper-parameter tunning, please refer to the in-text comments in AnnEconForecast/hpt_configs/template.py
for more information.
Documentations
AnnEconForecast/docs
Here is the documentation.
Training? Restoring? HPT? Run the Notebook
AnnEconForecast/notebooks
We used Jupyter notebooks to create an interactive environment for training, restoring and visualizing models in this project, each notebook contains detailed comments on how to use them.
Earlier models
ANNEF is a long-term project, but this repository is a new one. We have merged codes for earlier models into this repository and put them under this category.
Currently, they are not under active-development, but we might restart some of them in the future.
/keras_based
models built onkeras
/matlab_based
models built onMatLab