Traditional learning systems are trained in closed-world for a fixed number
of classes, and need pre-collected datasets in advance. However, new classes
often emerge in real-world applications and should be learned incrementally.
For example, in electronic commerce, new types of products appear daily, and in
a social media community, new topics emerge frequently. Under such
circumstances, incremental models should learn several new classes at a time
without forgetting. We find a strong correlation between old and new classes in
incremental learning, which can be applied to relate and facilitate different
learning stages mutually. As a result, we propose CO-transport for class
Incremental Learning (COIL), which learns to relate across incremental tasks
with the class-wise semantic relationship. In detail, co-transport has two
aspects: prospective transport tries to augment the old classifier with optimal
transported knowledge as fast model adaptation. Retrospective transport aims to
transport new class classifiers backward as old ones to overcome forgetting.
With these transports, COIL efficiently adapts to new tasks, and stably resists
forgetting. Experiments on benchmark and real-world multimedia datasets
validate the effectiveness of our proposed method.

CO-transport for Class Incremental Learning (COIL) is proposed as a method for incremental learning that utilizes class-wise semantic relationships and prospective and retrospective transport to adapt to new tasks and resist forgetting, as validated in experiments on benchmark and real-world multimedia datasets.