[初賽第二階段] 神經網路（2） [Preliminary Phase 2] Neural Networks (2)

Big Concept: Neural Networks Topic: From learning with examples to learning internal representations Bloom's Taxonomy: remember, understand Learning Goal: Deliver a clear lecture explaining artificial neural networks by starting from K-Nearest Neighbors (KNN) as a simple neural-network-like model. As a useful teaching analogy, explain that KNN can be viewed as a model whose hidden layer stores the training examples: each hidden unit compares the new input with one stored example, and the output is produced by voting or averaging over the most similar examples. Then introduce neural networks as a more advanced version of this idea. Instead of using the training examples themselves as hidden units, neural networks learn hidden units, weights, and internal representations from data. Explain the basic ideas of inputs, hidden layers, neurons, activations, weights, training, and outputs. Use real-world examples. Avoid advanced mathematics. Focus on how neural networks move from memorizing examples to learning useful representations for prediction. ( KNN is a useful bridge for understanding neural networks, but the main focus of the lecture should be understanding neural networks.) Key Learning Points: neural network; k-nearest neighbors; k-nearest neighbors as example-based learning; hidden layer as stored examples; learned representation