Here are my solutions for the course CS231N: Convolutional Neural Networks for Visual Recognition (Winter 2016). I've learnt a lot about deep learning especially CNN through courses materials and finished all assignments. However, even though I've passed all tests in the assignments, I'm not sure all the codes are exactlly right and efficient. Therefore I'm pleased if anyone can find some problems in my code and let me know.
Concreatlly, I've finished these assignments:
- Q1: k-Nearest Neighbor classifier: Implementing the kNN classifier.
- Q2: Training a Support Vector Machine: Implementing the SVM classifier.
- Q3: Implement a Softmax classifier: Implementing the Softmax classifier.
- Q4: Two-Layer Neural Network: The implementation of a two-layer neural network classifier.
- Q5: Higher Level Representations: Examine the improvements gained by using higher-level representations as opposed to using raw pixel values
- Q1: Fully-connected Neural Network: Implement fully-connected networks of arbitrary depth.
- Q2: Batch Normalization: Implement batch normalization, and use it to train deep fully-connected networks.
- Q3: Dropout: Implement Dropout and explore its effects on model generalization.
- Q4: ConvNet on CIFAR-10: Implement several new layers that are commonly used in convolutional networks and train a shallow convolutional network on CIFAR-10.
- Q1: Image Captioning with Vanilla RNNs: Implementation of an image captioning system on MS-COCO using vanilla recurrent networks.
- Q2: Image Captioning with LSTMs: Implementation of Long-Short Term Memory (LSTM) RNNs, and apply them to image captioning on MS-COCO.
- Q3: Image Gradients: Compute gradients with respect to the image, and use them to produce saliency maps and fooling images.
- Q4: Image Generation: Compute gradients with respect to the image, and generate class visualizations and implement feature inversion and DeepDream.
A big thank to all the members from the CS231N course. Thanks for your fantastic work and selfless contributions.