#First of all install and unzip ml-25m from ml-25m.zip from https://files.grouplens.org/datasets/movielens/
# Wide and Deep Learning For Recommender Systems
This is my PyTorch implementation of the paper [Wide & Deep Learning for Recommender Systems](https://arxiv.org/pdf/1606.07792.pdf) by Google Inc. (2016).
This model has been productionized and evaluated on [Google Play](https://play.google.com/store?hl=en_US), a commercial mobile app store with over one billion active users and over one million apps.
## Results
Here are the model hyper-parameters chosen:
- Number of Dense Embedding Dimensions used in the Deep Component = 16
- Number of Hidden Layers used in the Deep Component = 16
- Activation Function = Sigmoid
- Learning Rate = 0.001
- Batch Size = 512
- Weight Decay = 0.000001
- Optimizer Method = Adam
- Dropout Rate = 0.5
After being trained for 100 epochs, the model achieves **validation AUC = 0.7995** and **test AUC = 0.7991** with **runtime = 1h 12m 15s**
The results can be viewed at [this Weights & Biases link](https://app.wandb.ai/khanhnamle1994/multi_layer_perceptron_collaborative_filtering/runs/4jjdo87k). # Redes-Neurais