Computational neuroscience

 I follow Geoffrey Hinton's research because AI, a computer science field, needs cognitive psychologists' and computational neuroscientist's input. New algorithms in AI are increasingly inspired by fields relevant to brain research. One of the grand challenges in computing is for sure a radical increase in generalization and learning capabilities. Part of the problem is obviously related to training data sizes. The human brain, from early childhood, is exposed to a vast amount of data through our senses. There is also a massive amount of feedback and temporal relationships. Many machine learning applications are not trained this way, although some scientists emphasize long-term learning, e.g., Tom Mitchell. Bengio and Hinton are attacking the algorithmic problem, which is the second part of the challenge. The article in Quanta is an excellent overview of fundamental neural network algorithms. #machinelearning #ai #deeplearning #neuralnetworks #ann #backprop #backpropagation #data #algorithms #research #ml #datascience #evolutionary #evolutionaryalgorithms #neurosciences

https://www.quantamagazine.org/artificial-neural-nets-finally-yield-clues-to-how-brains-learn-20210218/