Hristo Vrigazov: Multi-task learning in the real world with Catalyst

Data Fest Online 2020 Catalyst Workshop Track Many real-world problems can be solved very computationally cheaply using multi-task learning - where we share weights between tasks and have task-specific layers only in the final few layers of the neural network. But the complexity of managing a multi-task learning project grows very quickly. Using Catalyst makes such a project much easier in the real world, and we will have a look at three different use-cases and how Catalyst was used to greatly ease the training, evaluation, interpretation and tuning of the results in a multi-task setting. Register and get access to the tracks: Join the community: