2019/12/17 by Geunseob Oh, Oh, Geunseob, Jean‐Sebastien Valois +1
Computer Science · Engineering · #Data Stream Mining Techniques #FOS: Computer and information sciences #Gaussian Processes and Bayesian Inference #Machine Learning (cs.LG) #Machine Learning (stat.ML) #Robotics (cs.RO) #Traffic Prediction and Management Techniques
paper · pdf · doi:10.48550/arxiv.1912.08111
openalex publication_date 2019/12/17 · openalex created_date 2025/10/10 · openalex updated_date 2026/07/28
We introduce Hyper-Conditioned Neural Autoregressive Flow (HCNAF); a powerful\nuniversal distribution approximator designed to model arbitrarily complex\nconditional probability density functions. HCNAF consists of a neural-net based\nconditional autoregressive flow (AF) and a hyper-network that can take large\nconditions in non-autoregressive fashion and outputs the network parameters of\nthe AF. Like other flow models, HCNAF performs exact likelihood inference. We\nconduct a number of density estimation tasks on toy experiments and MNIST to\ndemonstrate the effectiveness and attributes of HCNAF, including its\ngeneralization capability over unseen conditions and expressivity. Finally, we\nshow that HCNAF scales up to complex high-dimensional prediction problems of\nthe magnitude of self-driving and that HCNAF yields a state-of-the-art\nperformance in a public self-driving dataset.\n