Update README.md

README.md

@@ -1,10 +1,13 @@
-## Interpretive deep learning for hydrological understanding
+## Interpretive deep learning for investigating flooding mechanism
+[![DOI](https://img.shields.io/badge/DOI-10.5281/zenodo.4686106-blue.svg)](https://doi.org/10.5281/zenodo.4686106)
 
 - [Overview](#overview)
 - [Quick Start](#quick-start)
 
 ### Overview
-The code demonstrates the implementation of the interpretive LSTM proposed in paper "***Gaining process understanding from the black box: Uncovering flooding mechanisms through interpretive deep learning***" (submitted to a journal)
+The code demonstrates the implementation of using expected gradients and additive decomposition methods to interpret LSTM-based hydrological models, as proposed in the paper "***Uncovering flooding mechanisms across the contiguous United States through interpretive deep learning on representative catchments***" (submitted to a journal)
+
+**Note: some codes will be released once the paper is accepted.**
 
 Please refer to the file [LICENSE](/LICENSE) for the license governing this code.
 
@@ -23,14 +26,18 @@ The code was tested with Python 3.7. To use the code, please do:
    cd hydro-interpretive-dl
    ```
 
+> The study is implemented based on MOPEX (Model Parameter Estimation Experiment) dataset by [*NOAA National Weather Service*](https://www.nws.noaa.gov/ohd/mopex/mo_datasets.htm). 
+> One can download the data from the [official website](https://hydrology.nws.noaa.gov/pub/gcip/mopex/US_Data/) or [HydroShare](https://www.hydroshare.org/resource/99d5c1a238134ea6b8b767a65f440cb7/data/contents/MOPEX.zip) into `mopex`.
+
 2. Install dependencies ([conda](https://docs.conda.io/en/latest/miniconda.html) is recommended to manage packages):
 
    ```shell
 	conda create -n hydrodeepx
 	conda activate hydrodeepx
-	conda install -c conda-forge python=3.7 numpy=1.16.4 pandas scipy tensorflow=1.14 matplotlib jupyter h5py=2.10 shap tqdm
+	conda install -c conda-forge python=3.7 numpy=1.16.4 tensorflow=1.14 h5py=2.10 keras shap
+	conda install -c conda-forge pandas scipy matplotlib jupyter tqdm
    ```
+
+   Note for this implementation, `tensorflow v1.14` is recommended, though `tensorflow v2.x` may also work.
 
-3. Start `Jupyter Notebook` and run the `demo.ipynb` locally.
-
-> The study is implemented based on MOPEX (Model Parameter Estimation Experiment) dataset by [*NOAA National Weather Service*](https://www.nws.noaa.gov/ohd/mopex/mo_datasets.htm). One can download the data from the [official website](https://hydrology.nws.noaa.gov/pub/gcip/mopex/US_Data/) or [HydroShare](https://www.hydroshare.org/resource/99d5c1a238134ea6b8b767a65f440cb7/data/contents/MOPEX.zip) into `mopex`.
+3. Start `Jupyter Notebook` and run the `interpret_lstm.ipynb` locally.