Embeddings are a way to map one type of data into another space. Use embeddings to map different data types into the same space. Example: Temperature is numerical data. A description of deposits along a reactor surface is text data. Images of deposits are yet another type: image data.
Exploring embeddings is at the forefront of artificial intelligence. Embeddings have been used in natural language processing and computer vision models.
Begin by creating datasets that have pairs of text and images, or pairs of text and numerical process conditions (temperature, pressure, humidity, composition).
Define a loss function that optimizes embeddings. Elements that are paired should be close to each other in the embedded space.
Create an embedding model.
Use the embedding model to find images based on text, process conditions, or any variables you chose to incorporate in your training.
Use Naclai to work with a local engineer in Seattle, Washington; Pittsburgh, Pennsylvania; Albany, New York; and Boulder, Colorado.
Measurements in Chemical and Environmental Engineering