DrCod/Blood-Spectroscopy-Classification-using-Deep-Learning

My solution code to a machine learning problem from Zinid.Africa. # Blood-Spectroscopy-Classification-using-Deep-Learning A starter solution code(including the best final solution iteration) to a machine learning problem from Zindi.Africa. Find link to final notebook here Starter iteration utilized FFT(fast fourier transform) features to train a neural network architecture in a single/multi-input fashion. The ultimate goal of this project was to accurately identify the levels of different blood-based substrates from Near-Infrared Reflectance(NIR) spectral data. Full description of challenge and datasets acquisition can be accessed here # How to ran starter scripts in single/multi-input modes # single input: python train.py --train_csv Train.csv --test_csv Test.csv --use_threshold --use_smoothing --BATCH_SIZE 32 \ --EARLY_STOP --EPOCHS 100 --WEIGHT_DECAY 1e-6 --model_type single --model_name NN # multiple inputs: python train.py --train_csv Train.csv --test_csv Test.csv --use_threshold --use_smoothing --BATCH_SIZE 32 \ --EARLY_STOP --EPOCHS 100 --WEIGHT_DECAY 1e-6 --model_type double --model_name NN # run inference python make_submission.py --model_name NN --submission_dir path/to/submissions # Leaderboard Rank : 14/265 # Authors | Name | Zindi ID | Github ID | | ------------- |:-------------:| -----:| | Ronny Polle | @100i | @DrCod |