Deep Archetypal Analysis
This work provides a method for data-driven representation learning that reduces reliance on expert knowledge, with potential applications in biology and chemistry, though it is incremental as it builds upon existing linear Archetypal Analysis.
The paper tackles the problem of generating interpretable latent representations for high-dimensional datasets by extending linear Archetypal Analysis to a deep learning framework, enabling unsupervised and supervised applications such as identifying archetypal faces in CelebA and exploring chemical space for molecular design.
"Deep Archetypal Analysis" generates latent representations of high-dimensional datasets in terms of fractions of intuitively understandable basic entities called archetypes. The proposed method is an extension of linear "Archetypal Analysis" (AA), an unsupervised method to represent multivariate data points as sparse convex combinations of extremal elements of the dataset. Unlike the original formulation of AA, "Deep AA" can also handle side information and provides the ability for data-driven representation learning which reduces the dependence on expert knowledge. Our method is motivated by studies of evolutionary trade-offs in biology where archetypes are species highly adapted to a single task. Along these lines, we demonstrate that "Deep AA" also lends itself to the supervised exploration of chemical space, marking a distinct starting point for de novo molecular design. In the unsupervised setting we show how "Deep AA" is used on CelebA to identify archetypal faces. These can then be superimposed in order to generate new faces which inherit dominant traits of the archetypes they are based on.