The Allen Institute for Artificial Intelligence has organized a 4 month contest in Kaggle on question answering. The aim is to create a system which can correctly answer the questions from the 8th grade science exams of US schools (biology, chemistry, physics etc.). DeepHack Lab organized a scientific school + hackathon devoted to this contest in Moscow. Our team decided to use this opportunity to explore the deep learning techniques on question answering (although they seem to be far behind traditional systems). We tried to implement Dynamic memory networks described in a paper by A. Kumar et al. Here we report some preliminary results. In the next blog post we will describe the techniques we used to get to top 5% in the contest.
Few months ago Andrej Karpathy wrote a great blog post about recurrent neural networks. He explained how these networks work and implemented a character-level RNN language model which learns to generate Paul Graham essays, Shakespeare works, Wikipedia articles, LaTeX articles and even C++ code. He also released the code of the network on Github. Lots of people did experiments, like generating recipes, Bible or Irish folk music. We decided to test it on some legal texts in Armenian.
Recently TopCoder announced a contest to identify the spoken language in audio recordings. I decided to test how well deep convolutional networks will perform on this kind of data. In short I managed to get around 95% accuracy and finished at the 10th place. This post reveals all the details.
After watching the awesome video course by Hugo Larochelle on neural nets (more on this in the previous post) we decided to test our knowledge on some computer vision contest. We looked at Kaggle and the only active competition related to computer vision (except for the digit recognizer contest, for which lots of perfect out-of-the-box solutions exist) was the Diabetic retinopathy detection contest. This was probably quite hard to become our very first project, but nevertheless we decided to try. The team included Karen, Tigran, Hrayr, Narek (1st to 3rd year bachelor students) and me (PhD student). Long story short, we finished at the 82nd place out of 661 participants, and in this post I will describe in details what we did and what mistakes we made. All required files are on these 2 github repositories. We hope this will be interesting for those who just start to play with neural networks. Also we hope to get feedback from experts and other participants.
Who we are
We are a group of students from the department of Informatics and Applied Mathematics at Yerevan State University. In 2014, inspired by successes of neural nets in various fields, especially by GoogLeNet’s excellent performance in ImageNet 2014, we decided to dive into the topic of neural networks. We study calculus, combinatorics, graph theory, algebra and many other topics in the university but we learn nothing about machine learning. Just a few students take some ML courses from Coursera or elsewhere.