Very-Low Random Projection Maps
نویسندگان
چکیده
For Big Data analytics, working in low dimensionalities is beneficial for high performance. Instead of projecting onto a single low dimensionality, we examine, both analytically and empirically, the effects on the ‘learning utility’ of the original dataset when combining several very low-dimensional random projections. The embedding proposed exhibits many favorable traits to existing low-dimensional methodologies, such as low runtime and equivalent or better embedding quality.
منابع مشابه
Recovering the Optimal Solution by Dual Random Projection
Random projection has been widely used in data classification. It maps high-dimensional data into a low-dimensional subspace in order to reduce the computational cost in solving the related optimization problem. While previous studies are focused on analyzing the classification performance of using random projection, in this work, we consider the recovery problem, i.e., how to accurately recove...
متن کاملRandomness in Deconvolutional Networks for Visual Representation
We systematically study the deep representation of random weight CNN (convolutional neural network) using the DeCNN (deconvolutional neural network) architecture. We first fix the weights of an untrained CNN, and for each layer of its feature representation, we train a corresponding DeCNN to reconstruct the input image. As compared with the pre-trained CNN, the DeCNN trained on a random weight ...
متن کاملRandom Projection-Based Anderson-Darling Test for Random Fields
In this paper, we present the Anderson-Darling (AD) and Kolmogorov-Smirnov (KS) goodness of fit statistics for stationary and non-stationary random fields. Namely, we adopt an easy-to-apply method based on a random projection of a Hilbert-valued random field onto the real line R, and then, applying the well-known AD and KS goodness of fit tests. We conclude this paper by studying the behavior o...
متن کاملK-Dynamical Self Organizing Maps
Neural maps are a very popular class of unsupervised neural networks that project high-dimensional data of the input space onto a neuron position in a low-dimensional output space grid. It is desirable that the projection effectively preserves the structure of the data. In this paper we present a hybrid model called K-Dynamical Self Organizing Maps (KDSOM ) consisting of K Self Organizing Maps ...
متن کاملAdditive maps on C$^*$-algebras commuting with $|.|^k$ on normal elements
Let $mathcal {A} $ and $mathcal {B} $ be C$^*$-algebras. Assume that $mathcal {A}$ is of real rank zero and unital with unit $I$ and $k>0$ is a real number. It is shown that if $Phi:mathcal{A} tomathcal{B}$ is an additive map preserving $|cdot|^k$ for all normal elements; that is, $Phi(|A|^k)=|Phi(A)|^k $ for all normal elements $Ainmathcal A$, $Phi(I)$ is a projection, and there exists a posit...
متن کامل