Time Series with PyTorch

5

Simple Neural Architecture

“Deep learning is going to be able to do everything.”—Geoffrey Hinton

As we continue our journey into time series forecasting with PyTorch, we are ready to delve into neural networks and build our own. Compared to traditional statistical methods like generalized linear models, neural networks are more complex modeling structures — a complexity that grants them remarkable flexibility and power. This is not as daunting as it seems; neural architectures are, in reality, combinations of mathematical operations arranged in networks, and as we’ll see in later chapters, even these networks can be combined into greater structures. Think of these mathematical components as building blocks — like Lego pieces — that we can ...