Selective manipulation of data attributes using deep generative models is an\nactive area of research. In this paper, we present a novel method to structure\nthe latent space of a Variational Auto-Encoder (VAE) to encode different\ncontinuous-valued attributes explicitly. This is accomplished by using an\nattribute regularization loss which enforces a monotonic relationship between\nthe attribute values and the latent code of the dimension along which the\nattribute is to be encoded. Consequently, post-training, the model can be used\nto manipulate the attribute by simply changing the latent code of the\ncorresponding regularized dimension. The results obtained from several\nquantitative and qualitative experiments show that the proposed method leads to\ndisentangled and interpretable latent spaces that can be used to effectively\nmanipulate a wide range of data attributes spanning image and symbolic music\ndomains.\n