Högskolan i Skövde

Visual Analytics has brought forward many solutions to different tasks such as exploring topics, understanding user and customer behavior, comparing genomes, or detecting anomalies. Many of these solutions, if not most, are standalone applications with technological contributions which cannot be easily taken for: reuse in other domains, further improvement, benchmarking, or integration and deployment alongside other solutions. The latter can prove specially helpful for exploratory data analysis. This often leads researchers to re-implement solutions and thus to a suboptimal use of skills and resources. This paper discusses further the lack of off-the-shelf libraries for Visual Analytics, and proposes the creation of pluggable libraries on top of existing technologies such as Spark and Zeppelin. We provide an illustrative example of a pluggable, Visual Analytics library using these technologies.

We propose a taxonomy for classifying and describing papers which contribute to making Machine Learning (ML) techniques interactive and interpretable for users. The taxonomy is composed of six elements – Dataset, Optimizer, Model, Predictions, Evaluator and Goodness – where each can bemade available for user interpretation and interaction. We give definitions to the terms interpretable and interactive in the context of useroriented Machine Learning, describe the role of each of the elements in the taxonomy, and describe papers as seen through the lens of the proposed taxonomy.

This paper argues for the use of a topology learning algorithm, the Growing Neural Gas (GNG), for providing an overview of the structure of large and multidimensional datasets that can be used in exploratory data analysis. We introduce a generic, off-the-shelf library, Visual GNG, developed using the Big Data framework Apache Spark, which provides an incremental visualization of the GNG training process, and enables user-in-the-loop interactions where users can pause, resume or steer the computation by changing optimization parameters. Nine case studies were conducted with domain experts from different areas, each working on unique real-world datasets. The results show that Visual GNG contributes to understanding the distribution of multidimensional data; finding which features are relevant in such distribution; estimating the number of k clusters to be used in traditional clustering algorithms, such as K-means; and finding outliers.

This paper presents a model for the progressive visualization and exploration of the structure of large datasets. That is, an abstraction on different components and relations which provide means for constructing a visual representation of a dataset’s structure, with continuous system feedback and enabled user interactions for computational steering, in spite of size. In this context, the structure of a dataset is regarded as the distance or neighborhood relationships among its data points. Size, on the other hand, is defined in terms of the number of data points. To prove the validity of the model, a proof-of-concept was developed as a Visual Analytics library for Apache Zeppelin and Apache Spark. Moreover, nine user studies where carried in order to assess the usability of the library. The results from the user studies show that the library is useful for visualizing and understanding the emerging cluster patterns, for identifying relevant features, and for estimating the number of clusters. 

This article presents an empirical user study that compares eight multidimensional projection techniques for supporting the estimation of the number of clusters, k, embedded in six multidimensional data sets. The selection of the techniques was based on their intended design, or use, for visually encoding data structures, that is, neighborhood relations between data points or groups of data points in a data set. Concretely, we study: the difference between the estimates of k as given by participants when using different multidimensional projections; the accuracy of user estimations with respect to the number of labels in the data sets; the perceived usability of each multidimensional projection; whether user estimates disagree with k values given by a set of cluster quality measures; and whether there is a difference between experienced and novice users in terms of estimates and perceived usability. The results show that: dendrograms (from Ward's hierarchical clustering) are likely to lead to estimates of k that are different from those given with other multidimensional projections, while Star Coordinates and Radial Visualizations are likely to lead to similar estimates; t-Stochastic Neighbor Embedding is likely to lead to estimates which are closer to the number of labels in a data set; cluster quality measures are likely to produce estimates which are different from those given by users using Ward and t-Stochastic Neighbor Embedding; U-Matrices and reachability plots will likely have a low perceived usability; and there is no statistically significant difference between the answers of experienced and novice users. Moreover, as data dimensionality increases, cluster quality measures are likely to produce estimates which are different from those perceived by users using any of the assessed multidimensional projections. It is also apparent that the inherent complexity of a data set, as well as the capability of each visual technique to disclose such complexity, has an influence on the perceived usability.

As large datasets become more common, so becomes the necessity for exploratory approaches that allow iterative, trial-and-error analysis. Without such solutions, hypothesis testing and exploratory data analysis may become cumbersome due to long waiting times for feedback from computationally-intensive algorithms. This work presents a process model for progressive multidimensional projections (P-MDPs) that enables early feedback and user involvement in the process, complementing previous work by providing a lower level of abstraction and describing the specific elements that can be used to provide early systemfeed back, and those which can be enabled for user interaction. Additionally, we outline a set of design constraints that must be taken into account to ensure the usability of a solution regarding feedback time, visual cluttering, and the interactivity ofthe view. To address these constraints, we propose the use of incremental vector quantization (iVQ) as a core step within the process. To illustrate the feasibility of the model, and the usefulness of the proposed iVQ-based solution, we present a prototype that demonstrates how the different usability constraints can be accounted for, regardless of the size of a dataset.

The growing neural gas (GNG) is an unsupervised topology learning algorithm that models a data space through interconnected units that stand on the populated areas of that space. Its output is a graph that can be visually represented on a two-dimensional plane, and be used as means to disclose cluster patterns in datasets. GNG, however, creates highly connected graphs when trained on high dimensional data, which in turn leads to highly clutter representations that fail to disclose any meaningful patterns. Moreover, its sequential learning limits its potential for faster executions on local datasets, and, more importantly, its potential for training on distributed datasets while leveraging from the computational resources of the infrastructures in which they reside.

This paper presents two methods that improve GNG for the visualization of cluster patterns in large and high-dimensional datasets. The first one focuses on providing more meaningful and accurate cluster pattern representations of high-dimensional datasets, by avoiding connections that lead to high-dimensional graphs in the modeled topology, which may, in turn, lead to visual cluttering in 2D representations. The second method presented in this paper enables the use of GNG on big and distributed datasets with faster execution times, by modeling and merging separate parts of a dataset using the MapReduce model.

Quantitative and qualitative evaluations show that the first method leads to the creation of lower-dimensional graph structures, which in turn provide more accurate and meaningful cluster representations; and that the second method preserves the accuracy and meaning of the cluster representations while enabling its execution in distributed settings.