MLP, CNN and RNN
| Model | Connection Style | Best For | Weakness |
|---|---|---|---|
| MLP | Fully connected, dense layers | Tabular data, general-purpose learning | Ignores structure (order, spatial locality) |
| CNN | Convolutions (local filters) | Images, spatial data | Struggles with sequences without modification |
| RNN | Sequential recurrence | Time series, language | Hard with long dependencies, slow training |