Statistical Learning vs Neural Networks

STATISTICAL LEARNING Gentlemen, our learner overgeneralizes because the VC-Dimension of our Kernel is too high, Get some experts and minimze the structural risk in a new one. Rework our loss function, make the next kernel stable, unbiased and consider using a soft margin NEURAL NETWORKS STACK MORE LAYERS LAYERS LAYERS

A two-panel drawing compares statistical learning with neural networks. In the first panel, a stick figure proposes a technically sophisticated solution involving VC-dimension, kernels, loss functions, and soft margins; in the second, the neural-network solution is simply to stack more layers. The joke mocks the tendency to solve complex machine-learning problems with deep-learning brute force. EN: A two-panel drawing compares statistical learning with neural networks. In the first panel, a stick figure proposes a technically sophisticated solution involving VC-dimension, kernels, loss functions, and soft margins; in the second, the neural-network solution is simply to stack more layers. The joke mocks the tendency to solve complex machine-learning problems with deep-learning brute force.

• @profunctor_io / 2189 2019-05-13