Under the summer project at my institute IIT Kanpur "Model Zoo", we are learning and practising to implement and summarise various deep learning architectures and models.

Here is a list of the implemenations:

1. Resnet : Original Paper (Implemented)
   • Based on CNN, aimed to tackle the difficulty of training deeper networks without rise in training and testing error.
2. InfoGAN : Original Paper (Implemented)
   • Based on DCGan, aimed at disentangled representation so that network learns systematically various meaningful features of the input sample.
3. ALBERT : Original Paper (Under implementation)
   • Based on BERT, it addresses the GPU/TPU memory limitations and training complexities of the the former by proposing two parameter-reduction techniques, which lowers memory, increases training speed and even achieves greater accuracy.
4. Image Captioning : Original Paper (Implemented)
   • Based on a novel combination of CNN over images and bidirectional RNN (LSTM) over captions, using output of encoder CNN as embeddings for the decoder RNN to generate rich descriptions of image regions.
5. GloVe : Original Paper (Implemented)
   • GloVe is an unsupervised learning algorithm for obtaining vector representations for words. Training is performed on aggregated global word-word co-occurrence statistics from a corpus, and the resulting representations showcase interesting linear substructures of the word vector space.