🔑Generative models are probability distributions that can simulate data and generate new objects based on a given input.
🌟The structure of generative models often involves a combination of data and prior knowledge, such as architectural choices and loss functions.
🎨Generative models can be controlled using different signals, like captions for images or text in different languages.
⚙️Deep generative models, implemented using neural networks, are commonly used to build statistical data simulators.
🔍Generative models can also be used to query the model about the likelihood of certain data points being generated.