Which Tasks Should Be Learned Together in Multi-task Learning?

[Paper] [Video]

0) Motivation, Object and Related works:

Motivation:

Many computer vision applications require solving multiple tasks in real-time.
A neural network can be trained to solve multiple tasks simultaneously using multi-task learning.
Save computation at inference time as only a single network needs to be evaluated.
Task objectives can compete ==> leads to inferior overall performance.
Which tasks should and should not be learned together in one network when employing multi-task learning?

Objectives:

Study task cooperation and competition in several different learning settings
Propose a framework for assigning tasks to a few neural networks:
1. Cooperating tasks are computed by the same neural network
2. Competing tasks are computed by different networks.
Our framework offers:
1. A time-accuracy trade-off
2. Better accuracy.
3. Less inference time.

Contribution:

Related works:

Part 1: Study of Task Interactions:

Part 2: Task Grouping Framework

References:

Page updated

Google Sites

Report abuse