VISITOR: Visual Interactive State Sequence Exploration for Reinforcement Learning

Understanding the behavior of deep reinforcement learning agents is a crucial requirement throughout their development. Existing work has addressed the identification of observable behavioral patterns in state sequences or analysis of isolated internal representations; however, the overall decision-making of deep-learning RL agents remains opaque. To tackle this, we present VISITOR, a visual analytics system enabling the analysis of entire state sequences, the diagnosis of singular predictions, and the comparison between agents. A sequence embedding view enables the multiscale analysis of state sequences, utilizing custom embedding techniques for a stable spatialization of the observations and internal states. We provide multiple layers: (1) a state space embedding, highlighting different groups of states inside the state-action sequences, (2) a trajectory view, emphasizing decision points, (3) a network activation mapping, visualizing the relationship between observations and network activations, (4) a transition embedding, enabling the analysis of state-to-state transitions. The embedding view is accompanied by an interactive reward view that captures the temporal development of metrics, which can be linked directly to states in the embedding. Lastly, a model list allows for the quick comparison of models across multiple metrics. Annotations can be exported to communicate results to different audiences. Our two-stage evaluation with eight experts confirms the effectiveness in identifying states of interest, comparing the quality of policies, and reasoning about the internal decision-making processes.     

LINGO : Visually Debiasing Natural Language Instructions to Support Task Diversity

Cross-task generalization is a significant outcome that defines mastery in natural language understanding. Humans show a remarkable aptitude for this, and can solve many different types of tasks, given definitions in the form of textual instructions and a small set of examples. Recent work with pre-trained language models mimics this learning style: users can define and exemplify a task for the model to attempt as a series of natural language prompts or instructions. While prompting approaches have led to higher cross-task generalization compared to traditional supervised learning, analyzing ‘bias’ in the task instructions given to the model is a difficult problem, and has thus been relatively unexplored. For instance, are we truly modeling a task, or are we modeling a user's instructions? To help investigate this, we develop LINGO, a novel visual analytics interface that supports an effective, task-driven workflow to (1) help identify bias in natural language task instructions, (2) alter (or create) task instructions to reduce bias, and (3) evaluate pre-trained model performance on debiased task instructions. To robustly evaluate LINGO, we conduct a user study with both novice and expert instruction creators, over a dataset of 1,616 linguistic tasks and their natural language instructions, spanning 55 different languages. For both user groups, LINGO promotes the creation of more difficult tasks for pre-trained models, that contain higher linguistic diversity and lower instruction bias. We additionally discuss how the insights learned in developing and evaluating LINGO can aid in the design of future dashboards that aim to minimize the effort involved in prompt creation across multiple domains.     

DASS Good: Explainable Data Mining of Spatial Cohort Data

Developing applicable clinical machine learning models is a difficult task when the data includes spatial information, for example, radiation dose distributions across adjacent organs at risk. We describe the co-design of a modeling system, DASS, to support the hybrid human-machine development and validation of predictive models for estimating long-term toxicities related to radiotherapy doses in head and neck cancer patients. Developed in collaboration with domain experts in oncology and data mining, DASS incorporates human-in-the-loop visual steering, spatial data, and explainable AI to augment domain knowledge with automatic data mining. We demonstrate DASS with the development of two practical clinical stratification models and report feedback from domain experts. Finally, we describe the design lessons learned from this collaborative experience.     

Visual Analytics on Network Forgetting for Task-Incremental Learning

Task-incremental learning (Task-IL) aims to enable an intelligent agent to continuously accumulate knowledge from new learning tasks without catastrophically forgetting what it has learned in the past. It has drawn increasing attention in recent years, with many algorithms being proposed to mitigate neural network forgetting. However, none of the existing strategies is able to completely eliminate the issues. Moreover, explaining and fully understanding what knowledge and how it is being forgotten during the incremental learning process still remains under-explored. In this paper, we propose KnowledgeDrift, a visual analytics framework, to interpret the network forgetting with three objectives: (1) to identify when the network fails to memorize the past knowledge, (2) to visualize what information has been forgotten, and (3) to diagnose how knowledge attained in the new model interferes with the one learned in the past. Our analytical framework first identifies the occurrence of forgetting by tracking the task performance under the incremental learning process and then provides in-depth inspections of drifted information via various levels of data granularity. KnowledgeDrift allows analysts and model developers to enhance their understanding of network forgetting and compare the performance of different incremental learning algorithms. Three case studies are conducted in the paper to further provide insights and guidance for users to effectively diagnose catastrophic forgetting over time.     

DRLViz: Understanding Decisions and Memory in Deep Reinforcement Learning

We present DRLViz, a visual analytics interface to interpret the internal memory of an agent (e.g. a robot) trained using deep reinforcement learning. This memory is composed of large temporal vectors updated when the agent moves in an environment and is not trivial to understand due to the number of dimensions, dependencies to past vectors, spatial/temporal correlations, and co-correlation between dimensions. It is often referred to as a black box as only inputs (images) and outputs (actions) are intelligible for humans. Using DRLViz, experts are assisted to interpret decisions using memory reduction interactions, and to investigate the role of parts of the memory when errors have been made (e.g. wrong direction). We report on DRLViz applied in the context of video games simulators (ViZDoom) for a navigation scenario with item gathering tasks. We also report on experts evaluation using DRLViz, and applicability of DRLViz to other scenarios and navigation problems beyond simulation games, as well as its contribution to black box models interpretability and explain-ability in the field of visual analytics.     

Doppler Volume Rendering: A Dynamic,Piecewise Linear Spectral Representation for Visualizing Astrophysics Simulations

We present a novel approach for rendering volumetric data including the Doppler effect of light. Similar to the acoustic Doppler effect, which is caused by relative motion between a sound emitter and an observer, light waves also experience compression or expansion when emitter and observer exhibit relative motion. We account for this by employing spectral volume rendering in an emission–absorption model, with the volumetric matter moving according to an accompanying vector field, and emitting and attenuating light at wavelengths subject to the Doppler effect. By introducing a novel piecewise linearear representation of the involved light spectra, we achieve accurate volume rendering at interactive frame rates. We compare our technique to rendering with traditional point-based spectral representation, and demonstrate its utility using a simulation of galaxy formation.     

Human–Computer Collaboration for Visual Analytics: an Agent-based Framework

The visual analytics community has long aimed to understand users better and assist them in their analytic endeavors. As a result, numerous conceptual models of visual analytics aim to formalize common workflows, techniques, and goals leveraged by analysts. While many of the existing approaches are rich in detail, they each are specific to a particular aspect of the visual analytic process. Furthermore, with an ever-expanding array of novel artificial intelligence techniques and advances in visual analytic settings, existing conceptual models may not provide enough expressivity to bridge the two fields. In this work, we propose an agent-based conceptual model for the visual analytic process by drawing parallels from the artificial intelligence literature. We present three examples from the visual analytics literature as case studies and examine them in detail using our framework. Our simple yet robust framework unifies the visual analytic pipeline to enable researchers and practitioners to reason about scenarios that are becoming increasingly prominent in the field, namely mixed-initiative, guided, and collaborative analysis. Furthermore, it will allow us to characterize analysts, visual analytic settings, and guidance from the lenses of human agents, environments, and artificial agents, respectively.     

VA + Embeddings STAR: A State-of-the-Art Report on the Use of Embeddings in Visual Analytics

Over the past years, an increasing number of publications in information visualization, especially within the field of visual analytics, have mentioned the term “embedding” when describing the computational approach. Within this context, embeddings are usually (relatively) low-dimensional, distributed representations of various data types (such as texts or graphs), and since they have proven to be extremely useful for a variety of data analysis tasks across various disciplines and fields, they have become widely used. Existing visualization approaches aim to either support exploration and interpretation of the embedding space through visual representation and interaction, or aim to use embeddings as part of the computational pipeline for addressing downstream analytical tasks. To the best of our knowledge, this is the first survey that takes a detailed look at embedding methods through the lens of visual analytics, and the purpose of our survey article is to provide a systematic overview of the state of the art within the emerging field of embedding visualization. We design a categorization scheme for our approach, analyze the current research frontier based on peer-reviewed publications, and discuss existing trends, challenges, and potential research directions for using embeddings in the context of visual analytics. Furthermore, we provide an interactive survey browser for the collected and categorized survey data, which currently includes 122 entries that appeared between 2007 and 2023.     

Tac-Anticipator: Visual Analytics of Anticipation Behaviors in Table Tennis Matches

Anticipation skill is important for elite racquet sports players. Successful anticipation allows them to predict the actions of the opponent better and take early actions in matches. Existing studies of anticipation behaviors, largely based on the analysis of in-lab behaviors, failed to capture the characteristics of in-situ anticipation behaviors in real matches. This research proposes a data-driven approach for research on anticipation behaviors to gain more accurate and reliable insight into anticipation skills. Collaborating with domain experts in table tennis, we develop a complete solution that includes data collection, the development of a model to evaluate anticipation behaviors, and the design of a visual analytics system called Tac-Anticipator. Our case study reveals the strengths and weaknesses of top table tennis players' anticipation behaviors. In a word, our work enriches the research methods and guidelines for visual analytics of anticipation behaviors.     

Illustrative Motion Smoothing for Attention Guidance in Dynamic Visualizations

3D animations are an effective method to learn about complex dynamic phenomena, such as mesoscale biological processes. The animators' goals are to convey a sense of the scene's overall complexity while, at the same time, visually guiding the user through a story of subsequent events embedded in the chaotic environment. Animators use a variety of visual emphasis techniques to guide the observers' attention through the story, such as highlighting, halos – or by manipulating motion parameters of the scene. In this paper, we investigate the effect of smoothing the motion of contextual scene elements to attract attention to focus elements of the story exhibiting high-frequency motion. We conducted a crowdsourced study with 108 participants observing short animations with two illustrative motion smoothing strategies: geometric smoothing through noise reduction of contextual motion trajectories and visual smoothing through motion blur of context items. We investigated the observers' ability to follow the story as well as the effect of the techniques on speed perception in a molecular scene. Our results show that moderate motion blur significantly improves users' ability to follow the story. Geometric motion smoothing is less effective but increases the visual appeal of the animation. However, both techniques also slow down the perceived speed of the animation. We discuss the implications of these results and derive design guidelines for animators of complex dynamic visualizations.     

Boxer: Interactive Comparison of Classifier Results

Machine learning practitioners often compare the results of different classifiers to help select, diagnose and tune models. We present Boxer, a system to enable such comparison. Our system facilitates interactive exploration of the experimental results obtained by applying multiple classifiers to a common set of model inputs. The approach focuses on allowing the user to identify interesting subsets of training and testing instances and comparing performance of the classifiers on these subsets. The system couples standard visual designs with set algebra interactions and comparative elements. This allows the user to compose and coordinate views to specify subsets and assess classifier performance on them. The flexibility of these compositions allow the user to address a wide range of scenarios in developing and assessing classifiers. We demonstrate Boxer in use cases including model selection, tuning, fairness assessment, and data quality diagnosis.     

Quadratic-Attraction Subdivision

The idea of improving multi-sided piecewise polynomial surfaces, by explicitly prescribing their behavior at a central surface point, allows for decoupling shape finding from enforcing local smoothness constraints. Quadratic-Attraction Subdivision determines the completion of a quadratic expansion at the central point to attract a differentiable subdivision surface towards bounded curvature, with good shape also in-the-large.     

FlexEvent: going beyond Case-Centric Exploration and Analysis of Multivariate Event Sequences

In many domains, multivariate event sequence data is collected focused around an entity (the case). Typically, each event has multiple attributes, for example, in healthcare a patient has events such as hospitalization, medication, and surgery. In addition to the multivariate events, also the case (a specific attribute, e.g., patient) has associated multivariate data (e.g., age, gender, weight). Current work typically only visualizes one attribute per event (label) in the event sequences. As a consequence, events can only be explored from a predefined case-centric perspective. However, to find complex relations from multiple perspectives (e.g., from different case definitions, such as doctor), users also need an event- and attribute-centric perspective. In addition, support is needed to effortlessly switch between and within perspectives. To support such a rich exploration, we present FlexEvent: an exploration and analysis method that enables investigation beyond a fixed case-centric perspective. Based on an adaptation of existing visualization techniques, such as scatterplots and juxtaposed small multiples, we enable flexible switching between different perspectives to explore the multivariate event sequence data needed to answer multi-perspective hypotheses. We evaluated FlexEvent with three domain experts in two use cases with sleep disorder and neonatal ICU data that show our method facilitates experts in exploring and analyzing real-world multivariate sequence data from different perspectives.     

VA-TRAC: Geospatial Trajectory Analysis for Monitoring,Identification, and Verification in Fishing Vessel Operations

In order to ensure sustainability, fishing operations are governed by many rules and regulations that restrict the use of certain techniques and equipment, specify the species and size offish that can be harvested, and regulate commercial activities based on licensing schemes. As the world's second largest exporter offish and seafood products, Norway invests a significant amount of effort into maintaining natural ecosystem dynamics by ensuring compliance with its constantly evolving science-based regulatory body. This paper introduces VA-TRAC, a geovisual analytics application developed in collaboration with the Norwegian Directorate of Fisheries in order to address this complex task. Our approach uses automatic methods to identify possible catch operations based on fishing vessel trajectories, embedded in an interactive web-based visual interface used to explore the results, compare them with licensing information, and incorporate the analysts’ domain knowledge into the decision making process. We present a data and task analysis based on a close collaboration with domain experts, and the design and implementation of VA-TRAC to address the identified requirements.     

Hairy Slices II: Depth Cues for Visualizing 3D Streamlines Through Cutting Planes

Visualizing 3D vector fields is challenging because of occlusion problems and the difficulty of providing depth cues that adequately support the perception of direction of flow lines in 3D space. One of the depth cues that has proven most valuable for the perception of other kinds of 3D data, notably 3D networks and 3D point clouds, is structure-from-motion (also called the Kinetic Depth Effect); another powerful depth cue is stereoscopic viewing. We carried out an experiment of the perception of direction for short streamlines passing through a cutting plane. The conditions included viewing with and without structure-from-motion and with and without stereoscopic depth. Conditions also include comparing streamtubes to lines. The results show that for this particular task, stereo provided an effective depth cue, but structure-from-motion did not. Ringed streamtubes and streamcones provided good 3D direction information, even without stereoscopic viewing. We conclude with guidelines for viewing slices through vector fields.     

Fully Convolutional Graph Neural Networks for Parametric Virtual Try-On

We present a learning-based approach for virtual try-on applications based on a fully convolutional graph neural network. In contrast to existing data-driven models, which are trained for a specific garment or mesh topology, our fully convolutional model can cope with a large family of garments, represented as parametric predefined 2D panels with arbitrary mesh topology, including long dresses, shirts, and tight tops. Under the hood, our novel geometric deep learning approach learns to drape 3D garments by decoupling the three different sources of deformations that condition the fit of clothing: garment type, target body shape, and material. Specifically, we first learn a regressor that predicts the 3D drape of the input parametric garment when worn by a mean body shape. Then, after a mesh topology optimization step where we generate a sufficient level of detail for the input garment type, we further deform the mesh to reproduce deformations caused by the target body shape. Finally, we predict fine-scale details such as wrinkles that depend mostly on the garment material. We qualitatively and quantitatively demonstrate that our fully convolutional approach outperforms existing methods in terms of generalization capabilities and memory requirements, and therefore it opens the door to more general learning-based models for virtual try-on applications.