keras-team/autokeras

AutoKeras is a Python library that automates the process of building machine learning models. Normally, training a machine learning model requires choosing an architecture, tuning many settings by hand, and iterating through experiments. AutoKeras tries to handle that search automatically, so you can get a working model with far less manual effort. The library is built on top of Keras, which is itself a widely-used Python library for building neural networks. AutoKeras adds a layer on top of it that tries different model configurations on your behalf and returns the one that performs best on your data. The code example in the README shows the idea: you create a classifier, call fit with your training data, and call predict on new data, the same three steps you would use in any basic machine learning library, but the internal search for the right model structure happens automatically. The project comes from a research lab at Texas A&M University and has been published in an academic journal. It supports tasks like image classification, though the full list of supported task types is on the project's website rather than in the README itself. Installing it requires Python 3.7 or higher and TensorFlow 2.8.0 or higher, two common requirements in the Python machine learning ecosystem. The package is available via pip, the standard Python package installer. This tool is aimed at people who want to apply machine learning without spending a lot of time on model design. It is particularly useful for researchers or practitioners who want a baseline model quickly, or for beginners learning about machine learning who want results before they understand all the technical details.