Low-dimensional representation refers to the outcome of a dimension reduction process on high-dimensional data. The low-dimensional representation of the data is expected to retain as much information as possible from the high-dimensional data. Usually, there is a tradeoff between how low the dimension can be reduced and how much information can be retained.
Low-dimensional representations have many applications in the field of machine learning and deep learning. For example, one can use a low-dimensional representation of the data as a noise removal technique. Another example would be using a low-dimensional representation of the data as a feature-extraction technique. In general, if an appropriate low dimension is chosen, most of the information is retained from the original data. Using low-dimensional representations eases the learning task for machine learning algorithms without sacrificing too much model performance.
The C3 AI® Suite offers open source and commonly used dimensional reduction techniques as MLPipes. This helps the user to integrate low-dimensional representation of the data as a pre-processing step to other C3 components such as model training and model tuning. In addition, customized dimension reduction techniques can be constructed easily by extending the base MLPipe type within the C3 type system. These capabilities allow users to develop machine learning models with low-dimensional representation of the data seamlessly.