Implementation of Logistic Regression using Gradient Descent based on deeplearning.ai's course Neural Networks and Deep Learning (https://www.coursera.org/learn/neural-networks-deep-learning).
The logistic_regression.py file contains the implementation (in a scikit-learn-like interface), while the main.py file uses it to train a classifier using the data of Kaggle's Dogs vs Cats competition (https://www.kaggle.com/c/dogs-vs-cats/data), which contains 25.000 images of cats and dogs. To get image features, i used ResNet50 with Keras, of which i discarded the last layer and kept the last flatten layer, getting a 2048-dimensional vector per image.
Dataset split:
- 70% Training set (of which 10% is used for validation - the Validation set)
- 30% Test set
The model is trained using a learning rate of 0.01 and 1000 iterations. Below you can see the training and validation losses and accuracies (iterations in hundreds):
Accuracy in test set: 0.989333
And here are some predictions (class and probability) for 25 random test images:
