Deep Learning in a Structured Output Space

A large number of machine learning problems are considered as structured output problems in which the goal is to find the mapping function between an input vectors to a number of variables in the output side which are statistically correlated. Motivated by the advantages of simultaneous learning of these variables compared to learning them separately, many structured output models have been introduced. Decreasing the sample complexity, increasing the generalization ability and overcoming to noisy data are some of these benefits. So in the first step of this research we concentrate on one of classical but important problems in bioinformatics which is automatic protein function prediction. Results confirm that incorporating structural information helps to assign more detailed functionality to proteins. In addition to the accuracy and efficiency of a model, the ability of interpretation and explanation of its behavior is critical to apply it in real applications specifically in areas such as medicine. Therefore in the second step we focus on the problem of structured model interpretation and attempt to achieve a better understanding of the model behavior by exploring the structural information incorporated into the model to be explained.