Multi-Task_DeepLearning

This Repository contains the development of deep neural network models for binary classification using a set of tabular data. The deep learning framework used here is TensorFlow. Two models are developed: Two multi-task neural networks multitask_dnnclassifier_alternate.py and multitask_dnnclassifier_joined.py.

• For binary classification of two target variables simultaneously using one neural network.
• The deep learning models will extract information from both target variables in making prediction for each of the variables
• Models are being trained in two different fashion: 1) joined, 2) alternative.

Getting Started

Requirements

Requirements for TensorFlow Pandas Numpy Scikit-Learn

Installation

git clone https://github.com/nemaminejad/Multi-Task_DeepLearningn

Data

1. Collect your data, preprocess (Normalize, handle missing data,etc.)

2. Divide data into a training set, a validation set to use for finding best architecture and best performing model, a test set for final testing of model performance.

3. To use multi-task models:

• Place data in the form of multi_train.csv and multi_valid.csv
• Name your binary targets as var_1, var_2

5. for classification of two target variables simultaneously use the example script example_run_dnn_multitask.py

Under construction

Details of the model architectures, development and performance will be added soon.

Author

Nastaran Emaminejad

Find me in LinkedIn: https://www.linkedin.com/in/nastaran-emaminejad-791726137/

or Twitter: https://twitter.com/N_Emaminejad

Citation

If you found my work useful for your publications, please kindly cite this repository

"Nastaran Emaminejad, Multi-Task_DeepLearning, (2019), GitHub repository, https://github.com/nemaminejad/Multi-Task_DeepLearning "