Hands-On ROS for Robotics Programming

Deep learning

Deep learning is the most active research field in ML nowadays. The underlying model of this algorithm is a neural network whose way of working tries to mimic what the human brain does. Each neuron in the model performs a regression from its input with a special function, called sigmoid, that provides a sharp but continuous probability distribution of the output event. This function is the same as that of the probability function used in logistic regression, as described earlier. In this particular case of a neuron, if the resulting probability is greater than 50%, the neuron is activated and feeds another neuron or neurons downstream. If lower than 50%, the neuron is not active and hence it has negligible influence downstream. ...