LUNAR: Cellular automata for drifting data streams

With the advent of fast data streams, real-time machine learning has become a challenging task, demanding many processing resources. In addition, they can be affected by concept drift effect, which methods have to detect changes in distribution and adapt these evolving conditions. Several emerging paradigms such as so-called Smart Dust, Utility Fog, or Swarm Robotics are need for efficient scalable solutions scenarios, where usually computing resources constrained. Cellular automata, low-bias robust-to-noise pattern recognition with competitive classification performance, meet requirements imposed aforementioned mainly due their simplicity parallel nature. this work we propose LUNAR , streamified version cellular automata devised successfully requirements. is able act real incremental learner while adapting drifting Furthermore, highly interpretable, its structure represents directly mapping between feature space labels predicted. Extensive simulations synthetic will provide evidence behavior terms performance when compared long-established successful online methods.