In this example, the GlobalAveragePooling1D reduces the sequence to a 20-dimensional vector. The subsequent Dropout layer operates on exactly .
With nearly 30 seasons of content, it can be hard to know where to jump in. Here are a few fan-favorite starting points: Genre / Vibe Description Fantasy High Teen Comedy / Fantasy dropout dimension 20
Research continues to evolve. Recent papers on and Concrete Dropout allow the dropout rate itself to be learned per dimension. For a dropout dimension 20 scenario, this means the network could learn to drop certain features (e.g., positions 5, 12, and 18) more aggressively than others. In this example
import numpy as np