Task Discovery: Finding the Tasks that Neural Networks Generalize on

When developing deep learning models, we usually decide what task we want to solve then search for a model that generalizes well on the task. An intriguing question would be: what if, instead of fixing the task and searching in the model space, we fix the model and search in the task space? Can we find tasks that the model generalizes on? How do they look, or do they indicate anything? These are the questions we address in this paper. We propose a task discovery framework that automatically finds examples of such tasks via optimizing a generalization-based quantity called agreement score. We demonstrate that one set of images can give rise to many tasks on which neural networks generalize well. These tasks are a reflection of the inductive biases of the learning framework and the statistical patterns present in the data, thus they can make a useful tool for analysing the neural networks and their biases. As an example, we show that the discovered tasks can be used to automatically create adversarial train-test splits which make a model fail at test time, without changing the pixels or labels, but by only selecting how the datapoints should be split between the train and test sets. We end with a discussion on human-interpretability of the discovered tasks.

Ask Me Even More: Dynamic Memory Tensor Networks

We examine Memory Networks for the task of question answering (QA), under common real world scenario where training examples are scarce and under weakly supervised scenario, that is only extrinsic labels are available for training. We propose extensions for the Dynamic Memory Network (DMN), specifically within the attention mechanism, we call the resulting Neural Architecture as Dynamic Memory Tensor Network (DMTN). Ultimately, we see that our proposed extensions results in over 80% improvement in the number of task passed against the baselined standard DMN and 20% more task passed compared to state-of-the-art End-to-End Memory Network for Facebook's single task weakly trained 1K bAbi dataset.