Learning to perform addition operation using a fully-connected/dense neural network.
Classification on a Quickdraw dataset. Used a Inception module ConvNet coupled with SNN(Self-Normalising-Network) propety using SELU(Scaled Exponential Linear Units) as activations. Trained the inception modules from scratch.
Sentiment Analysis on the imdb dataset. Trained both word embedding layer as well as a layer of GRU.
CNN+BiRNN on noisy speech. Future work: GANs
Simple Conv and Dense layers works well. However, unable to generalise to different datasets. Future work : Explore Transfer learning to small datasets from other distributions.