concept conclusion in category deep learning
appears as: conclusion, conclusion, conclusions

This is an excerpt from Manning's book Deep Reinforcement Learning in Action.
Different neural network architectures such as convolutional, recurrent, or attention have different inductive biases that can improve learning when those biases are accurate. Inductive reasoning is when you observe some data and infer a more general pattern or rule from it. Deductive reasoning is what we do in mathematics when we start with some premises and by following logical rules assumed to be true, we can make a conclusion with certainty.
For example, the syllogism “All planets are round. Earth is a planet. Therefore, the Earth is round” is a form of deductive reasoning. There is no uncertainty about the conclusion if we assume the premises to be true.