ABSTRACT
A wide cross-section of emotionally intelligent robot algorithms is described in terms of their working mechanisms and advantages and drawbacks. Algorithms discussed include hidden Markov models (statistical models that analyze sequences of data), self-organizing maps (artificial neural networks to project high-dimensional data onto a low-dimensional grid), support vector machines (supervised learning algorithms for classification tasks), convolutional neural networks (deep learning architectures for image recognition and analysis), decision trees (tree-like structures for classification), natural language processing (analyzing the sentiment of spoken or written language to interpret the emotional tone of a conversation and adjust responses accordingly) and reinforcement learning (machine learning paradigms in which an agent learns by interacting with its environment).
