Home

This is summary of final project for my Artificial Neural Networks course, which I took last semester. A few students (including me) were suggested to learn more about Long Short Term Memory networks and do a little research on them, as a final task. I got interested and decided to take up the challenge.

#Introduction Long Short Term Memory (or LSTM) network is a type of Recurrent Neural Networks, where we introduce memory cell. Check out this article for excellent and very illustrative introduction to LSTM.

#Problem I worked on adding problem (tackled here or here), where input is a sequence of pairs. First component is a real number from the interval [0, 1] and the second component is 0 or 1. Exactly two elements are marked with 1. Desired output is the sum of those elements marked with ones.

My research is based on code provided by my professor, Jan Chorowski.

#Goals The ultimate goal would be getting rid of curriculum. A desire is to have a framework that we can use with problems where we cannot introduce curriculum learning.

I've made an attempt to achieve this as well as improve curriculum model.