Iterative Proportional Fitting 
The iterative proportional fitting procedure (IPFP, also known as biproportional fitting in statistics, RAS algorithm in economics and matrix ranking or matrix scaling in computer science) is an iterative algorithm for estimating cell values of a contingency table such that the marginal totals remain fixed and the estimated table decomposes into an outer product. …
Missing View Imputation with Generative Adversarial Networks (VIGAN)
In an era where big data is becoming the norm, we are becoming less concerned with the quantity of the data for our models, but rather the quality. With such large amounts of data collected from multiple heterogeneous sources comes the associated problems, often missing views. As most models could not handle whole view missing problem, it brings up a significant challenge when conducting any multi-view analysis, especially when used in the context of very large and heterogeneous datasets. However if dealt with properly, joint learning from these complementary sources can be advantageous. In this work, we present a method for imputing missing views based on generative adversarial networks called VIGAN which combines cross-domain relations given unpaired data with multi-view relations given paired data. In our model, VIGAN first learns bidirectional mapping between view X and view Y using a cycle-consistent adversarial network. Moreover, we incorporate a denoising multimodal autoencoder to refine the initial approximation by making use of the joint representation. Empirical results give evidence indicating VIGAN offers competitive results compared to other methods on both numeric and image data. …
Local Linear Forest
Random forests are a powerful method for non-parametric regression, but are limited in their ability to fit smooth signals, and can show poor predictive performance in the presence of strong, smooth effects. Taking the perspective of random forests as an adaptive kernel method, we pair the forest kernel with a local linear regression adjustment to better capture smoothness. The resulting procedure, local linear forests, enables us to improve on asymptotic rates of convergence for random forests with smooth signals, and provides substantial gains in accuracy on both real and simulated data. …
Like this:
Like Loading…
Related