A Deep Incremental Boltzmann Machine for Modeling Context in Robots
This work addresses context modeling for robots in adaptive environments, but it is incremental as it builds upon existing methods.
The paper tackles the problem of modeling context in robots by proposing an incremental deep model based on Restricted Boltzmann Machines that adapts to new scenes by adding contexts or layers, achieving performance on par or better than existing models on a scene classification benchmark.
Context is an essential capability for robots that are to be as adaptive as possible in challenging environments. Although there are many context modeling efforts, they assume a fixed structure and number of contexts. In this paper, we propose an incremental deep model that extends Restricted Boltzmann Machines. Our model gets one scene at a time, and gradually extends the contextual model when necessary, either by adding a new context or a new context layer to form a hierarchy. We show on a scene classification benchmark that our method converges to a good estimate of the contexts of the scenes, and performs better or on-par on several tasks compared to other incremental models or non-incremental models.