Title: LSTMs in Python (Keras and Lasagne

Speaker: Saurav Gupta

Abstract: Long Short Term Memory Neural Networks(LSTMs) are a variant of Recurrent Neural Networks that have worked well for a variety of tasks such as speech recognition, machine translation, language modeling and the like. We'll look at an example of training a language model with Keras. And we'll also implement custom layers in Lasagne with an example of Highway Networks [1].

[1] http://arxiv.org/abs/1505.00387