Communicative Learning: A Unified Learning Formalism

In this article, we propose a communicative learning (CL) formalism that unifies existing machine paradigms, such as passive learning, active algorithmic teaching, and so forth, facilitates the development of new methods. Arising from human cooperative communication, poses process combines pedagogy with burgeoning field learning. The pedagogical insight adoption alternative information sources in besides randomly sampled data, intentional messages given by helpful teacher. More specifically, CL, teacher student exchange each other collaboratively to transmit acquire certain knowledge. Each agent has mind, which includes agent’s knowledge, utility, mental dynamics. To establish effective also needs an estimation its partner’s mind. We define expressive representations formulation sufficient for recursive modeling, endows CL human-comparable efficiency. demonstrate application several prototypical collaboration tasks illustrate allows protocols go beyond Shannon’s communication limit. Finally, present our contribution foundations putting forth hierarchies defining halting problem