Multi-task Overview
Last updated
Last updated
Hard parameter sharing greatly reduces the risk of overfitting.The more tasks we are learning simultaneously, the more our model has to find a representation that captures all of the tasks and the less is our chance of overfitting on our original task.
Hard parameter sharing greatly reduces the risk of overfitting. In fact, [7] showed that the risk of overfitting the shared parameters is an order N -- where N is the number of tasks -- smaller than overfitting the task-specific parameters, i.e. the output layers. This makes sense intuitively: The more tasks we are learning simultaneously, the more our model has to find a representation that captures all of the tasks and the less is our chance of overfitting on our original task.
In soft parameter sharing on the other hand, each task has its own model with its own parameters. The distance between the parameters of the model is then regularized in order to encourage the parameters to be similar. [8] for instance use the ℓ2ℓ2 norm for regularization, while [9] use the trace norm.
