tasks.txt

- Implement neural network properly in Keras
- Benchmark LAR vs LAR+NNet vs LAR+eeRTree performance
- Write paper comparing SL2P vs LAR vs LAR+NNet estimates for biophysical variables
- Read Najib's ALR paper and MATLAB code
- Implement cross validation to determine when to stop LARs
    - 10 blocks of input data, determine RMSE for each, min RMSE and STDEV of RMSEs
    - Find min RMSE of all iterations, add stdev of RMSEs to it
    - Go to earliest iteration with min RMSE of 10 blocks under the above value
- Implement test criteria in stanford paper to determine iteration of LARs to stop on
- Implement trimming in python
- Implement orthogonal sampling for cross validation
- Implement ALR using python and GEE
- Test ALR using dataset provided by Najib

- Cross-validation with LARs
- Random forests on server side
- Implement SL2P in normalized predictions, with quality layer and error network
- Implement quality layer and error network in ALR, oversamples


- python
- ipython
- jupyter notebook/lab
- numpy
- scipy
- matplotlib
- pandas
- sympy
- cython
- ipyparallel
- scikit-learn / scikit-image
- tensorflow, keras
- pytorch

- R