LoCH: A neighborhood-based multidimensional projection technique for high-dimensional sparse spaces

Fadel, Samuel G.; Fatore, Francisco M.; Duarte, Felipe Simões Lage Gomes; Paulovich, Fernando V.

doi:10.1016/j.neucom.2014.07.071

Cited by 31 publications

(56 citation statements)

References 41 publications

Supporting

Mentioning

Contrasting

Unclassified

Order By: Relevance

“…Hybrid landmark techniques embed landmarks with non‐linear dimensionality reduction techniques based on high‐dimensional descriptors of the landmarks derived from the original data. The complete dataset is then embedded using different interpolation schemes [FFDP15, JPC*11, PNML08, PSN10, PdRDK99, dST04, PEP*11]. This approach is widely used by the visualization community due to its fast computation, making it ideal for interactive systems.…”

Section: Related Workmentioning

confidence: 99%

Hierarchical Stochastic Neighbor Embedding

Pezzotti

Höllt

Lelieveldt

et al. 2016

Computer Graphics Forum

135

147

View full text Add to dashboard Cite

In recent years, dimensionality-reduction techniques have been developed and are widely used for hypothesis generation in Exploratory Data Analysis. However, these techniques are confronted with overcoming the trade-off between computation time and the quality of the provided dimensionality reduction. In this work, we address this limitation, by introducing Hierarchical Stochastic Neighbor Embedding (Hierarchical-SNE). Using a hierarchical representation of the data, we incorporate the wellknown mantra of Overview-First, Details-On-Demand in non-linear dimensionality reduction. First, the analysis shows an embedding, that reveals only the dominant structures in the data (Overview). Then, by selecting structures that are visible in the overview, the user can filter the data and drill down in the hierarchy. While the user descends into the hierarchy, detailed visualizations of the high-dimensional structures will lead to new insights. In this paper, we explain how Hierarchical-SNE scales to the analysis of big datasets. In addition, we show its application potential in the visualization of Deep-Learning architectures and the analysis of hyperspectral images.

show abstract

Section: Related Workmentioning

confidence: 99%

Hierarchical Stochastic Neighbor Embedding

Pezzotti

Höllt

Lelieveldt

et al. 2016

Computer Graphics Forum

135

147

View full text Add to dashboard Cite

show abstract

“…For example, the Burnes-Hut trees were used for approximation of forces from distant particles in t-SNE realizaton of MDS [4]. Another approximation is used in LoCH algorithm [10], which seeks to place each point x i close to the convex hull of its nearest neighbors in X. All of these approximate methods have computational comlexity lower than O(M 2 ) (e.g., O(M log M ) for BH-SNE) and are really very efficient.…”

Section: Listingmentioning

confidence: 99%

“…As shown in [4,5], visualization of large data consisting of 10 5 + of objects requires approximated versions of MDS. They can be developed by limiting the number of computed distances, e.g., via random sampling [8,9], landmark particles [8], core points selection, hierarchical clustering and k-NN interpolation [10,11], or by using more sophisticated thinning or approximation procedures such as: deep belief networks (DBN) [12], Barnes-Hut-SNE [4], Q-SNE [5] or LoCH [10]. There are also many parallel realization of approximated versions of MDS including such the solvers as SMA-COF [13], GLIMMER [11] and SUBSET [8].…”

Section: Introductionmentioning

confidence: 99%

Visual Exploration of Data with Multithread MIC Computer Architectures

Pawliczek

Dzwinel

Yuen

2015

Artificial Intelligence and Soft Computing

View full text Add to dashboard Cite

Abstract. Knowledge mining from immense datasets requires fast, reliable and affordable tools for their visual and interactive exploration. Multidimensional scaling (MDS) is a good candidate for embedding of high-dimensional data into visually perceived 2-D and 3-D spaces. We focus here on the way to increase the computational performance of MDS in the context of interactive, hierarchical, visualization of big data. To this end we propose a parallel implementation of MDS on the modern Intel Many Integrated Core Architecture (MIC). We compare the timings obtained for MIC architecture to GPU and standard multi-core CPU implementations of MDS. We conclude that despite 30-40% lower computational performance comparing to GPU/CUDA tuned MDS codes, the MIC solution is still competitive due to dramatically shorter code production and tuning time. The integration of MIC with CPU will make this architecture very competitive with more volatile on technological changes GPU solutions.

show abstract

“…Conceptually, this technique tries to combine local and global neighborhood preservation. LoCH [30] attempted something similar by first projecting the control points which are clustering centers and then performing an iterative approximation in order to place each data point close to the convex hull of its nearest neighbors seeking to maintain small neighborhood structures. However, LoCH does not take into account interactive scenarios but focuses on where to place the neighboring points on the 2D space.…”

mentioning

confidence: 99%

Interactive dimensionality reduction using similarity projections

Spathis

Passalis

Tefas

2019

Knowledge-Based Systems

View full text Add to dashboard Cite

Recent advances in machine learning allow us to analyze and describe the content of high-dimensional data like text, audio, images or other signals. In order to visualize that data in 2D or 3D, usually Dimensionality Reduction (DR) techniques are employed. Most of these techniques, e.g., PCA or t-SNE, produce static projections without taking into account corrections from humans or other data exploration scenarios. In this work, we propose the interactive Similarity Projection (iSP), a novel interactive DR framework based on similarity embeddings, where we form a differentiable objective based on the user interactions and perform learning using gradient descent, with an end-to-end trainable architecture. Two interaction scenarios are evaluated. First, a common methodology in multidimensional projection is to project a subset of data, arrange them in classes or clusters, and project the rest unseen dataset based on that manipulation, in a kind of semi-supervised interpolation. We report results that outperform competitive baselines in a wide range of metrics and datasets. Second, we explore the * Corresponding author. Work done while at the Aristotle University of Thessaloniki.

show abstract

LoCH: A neighborhood-based multidimensional projection technique for high-dimensional sparse spaces

Cited by 31 publications

References 41 publications

Hierarchical Stochastic Neighbor Embedding

Hierarchical Stochastic Neighbor Embedding

Visual Exploration of Data with Multithread MIC Computer Architectures

Interactive dimensionality reduction using similarity projections

Contact Info

Product

Resources

About