Deep learning has had a lot of success in all sorts of fields. Thing is, that often means it's tough to figure out how it works. It also depends on massive amounts of data purely data-driven methods can skip over key ideas in science work. This work checks out Theory-Guided Neural Networks, or TGNNs. It's a new setup that weaves in physical laws and scientific smarts. Right into the network's structure and training process. We run through ways to blend in those theoretical bits. Hybrid models have come up. Regularization tricks too. And physics-informed neural networks, the PINNs. There is a case study with a public dataset. It shows how handy TGNNs are. They perform better in ways that matter.