LSTMVis: A Tool for Visual Analysis of Hidden State Dynamics in Recurrent Neural Networks

Visual explainable artificial intelligence for graph-based visual question answering and scene graph curation

This study presents a novel visualization approach to explainable artificial intelligence for graph-based visual question answering (VQA) systems. The method focuses on identifying false answer predictions by the model and offers users the opportunity to directly correct mistakes in the input space, thus facilitating dataset curation. The decision-making process of the model is demonstrated by highlighting certain internal states of a graph neural network (GNN). The proposed system is built on top of a GraphVQA framework that implements various GNN-based models for VQA trained on the GQA dataset. The authors evaluated their tool through the demonstration of identified use cases, quantitative measures, and a user study conducted with experts from machine learning, visualization, and natural language processing domains. The authors' findings highlight the prominence of their implemented features in supporting the users with incorrect prediction identification and identifying the underlying issues. Additionally, their approach is easily extendable to similar models aiming at graph-based question answering.

Towards Dataset-Scale and Feature-Oriented Evaluation of Text Summarization in Large Language Model Prompts

Recent advancements in Large Language Models (LLMs) and Prompt Engineering have made chatbot customization more accessible, significantly reducing barriers to tasks that previously required programming skills. However, prompt evaluation, especially at the dataset scale, remains complex due to the need to assess prompts across thousands of test instances within a dataset. Our study, based on a comprehensive literature review and pilot study, summarized five critical challenges in prompt evaluation. In response, we introduce a feature-oriented workflow for systematic prompt evaluation. In the context of text summarization, our workflow advocates evaluation with summary characteristics (feature metrics) such as complexity, formality, or naturalness, instead of using traditional quality metrics like ROUGE. This design choice enables a more user-friendly evaluation of prompts, as it guides users in sorting through the ambiguity inherent in natural language. To support this workflow, we introduce Awesum, a visual analytics system that facilitates identifying optimal prompt refinements for text summarization through interactive visualizations, featuring a novel Prompt Comparator design that employs a BubbleSet-inspired design enhanced by dimensionality reduction techniques. We evaluate the effectiveness and general applicability of the system with practitioners from various domains and found that (1) our design helps overcome the learning curve for non-technical people to conduct a systematic evaluation of summarization prompts, and (2) our feature-oriented workflow has the potential to generalize to other NLG and image-generation tasks. For future works, we advocate moving towards feature-oriented evaluation of LLM prompts and discuss unsolved challenges in terms of human-agent interaction.

AdversaFlow: Visual Red Teaming for Large Language Models with Multi-Level Adversarial Flow

Large Language Models (LLMs) are powerful but also raise significant security concerns, particularly regarding the harm they can cause, such as generating fake news that manipulates public opinion on social media and providing responses to unethical activities. Traditional red teaming approaches for identifying AI vulnerabilities rely on manual prompt construction and expertise. This paper introduces AdversaFlow, a novel visual analytics system designed to enhance LLM security against adversarial attacks through human-AI collaboration. AdversaFlow involves adversarial training between a target model and a red model, featuring unique multi-level adversarial flow and fluctuation path visualizations. These features provide insights into adversarial dynamics and LLM robustness, enabling experts to identify and mitigate vulnerabilities effectively. We present quantitative evaluations and case studies validating our system's utility and offering insights for future AI security solutions. Our method can enhance LLM security, supporting downstream scenarios like social media regulation by enabling more effective detection, monitoring, and mitigation of harmful content and behaviors.

Opening the AI Black Box: Distilling Machine-Learned Algorithms into Code

Can we turn AI black boxes into code? Although this mission sounds extremely challenging, we show that it is not entirely impossible by presenting a proof-of-concept method, MIPS, that can synthesize programs based on the automated mechanistic interpretability of neural networks trained to perform the desired task, auto-distilling the learned algorithm into Python code. We test MIPS on a benchmark of 62 algorithmic tasks that can be learned by an RNN and find it highly complementary to GPT-4: MIPS solves 32 of them, including 13 that are not solved by GPT-4 (which also solves 30). MIPS uses an integer autoencoder to convert the RNN into a finite state machine, then applies Boolean or integer symbolic regression to capture the learned algorithm. As opposed to large language models, this program synthesis technique makes no use of (and is therefore not limited by) human training data such as algorithms and code from GitHub. We discuss opportunities and challenges for scaling up this approach to make machine-learned models more interpretable and trustworthy.

An In-Situ Visual Analytics Framework for Deep Neural Networks

The past decade has witnessed the superior power of deep neural networks (DNNs) in applications across various domains. However, training a high-quality DNN remains a non-trivial task due to its massive number of parameters. Visualization has shown great potential in addressing this situation, as evidenced by numerous recent visualization works that aid in DNN training and interpretation. These works commonly employ a strategy of logging training-related data and conducting post-hoc analysis. Based on the results of offline analysis, the model can be further trained or fine-tuned. This strategy, however, does not cope with the increasing complexity of DNNs, because (1) the time-series data collected over the training are usually too large to be stored entirely; (2) the huge I/O overhead significantly impacts the training efficiency; (3) post-hoc analysis does not allow rapid human-interventions (e.g., stop training with improper hyper-parameter settings to save computational resources). To address these challenges, we propose an in-situ visualization and analysis framework for the training of DNNs. Specifically, we employ feature extraction algorithms to reduce the size of training-related data in-situ and use the reduced data for real-time visual analytics. The states of model training are disclosed to model designers in real-time, enabling human interventions on demand to steer the training. Through concrete case studies, we demonstrate how our in-situ framework helps deep learning experts optimize DNNs and improve their analysis efficiency.

Applications of Long Short-Term Memory (LSTM) Networks in Polymeric Sciences: A Review

This review explores the application of Long Short-Term Memory (LSTM) networks, a specialized type of recurrent neural network (RNN), in the field of polymeric sciences. LSTM networks have shown notable effectiveness in modeling sequential data and predicting time-series outcomes, which are essential for understanding complex molecular structures and dynamic processes in polymers. This review delves into the use of LSTM models for predicting polymer properties, monitoring polymerization processes, and evaluating the degradation and mechanical performance of polymers. Additionally, it addresses the challenges related to data availability and interpretability. Through various case studies and comparative analyses, the review demonstrates the effectiveness of LSTM networks in different polymer science applications. Future directions are also discussed, with an emphasis on real-time applications and the need for interdisciplinary collaboration. The goal of this review is to connect advanced machine learning (ML) techniques with polymer science, thereby promoting innovation and improving predictive capabilities in the field.

A Visual Analytics Conceptual Framework for Explorable and Steerable Partial Dependence Analysis

Machine learning techniques are a driving force for research in various fields, from credit card fraud detection to stock analysis. Recently, a growing interest in increasing human involvement has emerged, with the primary goal of improving the interpretability of machine learning models. Among different techniques, Partial Dependence Plots (PDP) represent one of the main model-agnostic approaches for interpreting how the features influence the prediction of a machine learning model. However, its limitations (i.e., visual interpretation, aggregation of heterogeneous effects, inaccuracy, and computability) could complicate or misdirect the analysis. Moreover, the resulting combinatorial space can be challenging to explore both computationally and cognitively when analyzing the effects of more features at the same time. This article proposes a conceptual framework that enables effective analysis workflows, mitigating state-of-the-art limitations. The proposed framework allows for exploring and refining computed partial dependences, observing incrementally accurate results, and steering the computation of new partial dependences on user-selected subspaces of the combinatorial and intractable space. With this approach, the user can save both computational and cognitive costs, in contrast with the standard monolithic approach that computes all the possible combinations of features on all their domains in batch. The framework is the result of a careful design process involving experts' knowledge during its validation and informed the development of a prototype, W4SP1, that demonstrates its applicability traversing its different paths. A case study shows the advantages of the proposed approach.

IdMotif: An Interactive Motif Identification in Protein Sequences

This article presents a visual analytics framework, idMotif, to support domain experts in identifying motifs in protein sequences. A motif is a short sequence of amino acids usually associated with distinct functions of a protein, and identifying similar motifs in protein sequences helps us to predict certain types of disease or infection. idMotif can be used to explore, analyze, and visualize such motifs in protein sequences. We introduce a deep-learning-based method for grouping protein sequences and allow users to discover motif candidates of protein groups based on local explanations of the decision of a deep-learning model. idMotif provides several interactive linked views for between and within protein cluster/group and sequence analysis. Through a case study and experts' feedback, we demonstrate how the framework helps domain experts analyze protein sequences and motif identification.

Visualizing and Comparing Machine Learning Predictions to Improve Human-AI Teaming on the Example of Cell Lineage

We visualize the predictions of multiple machine learning models to help biologists as they interactively make decisions about cell lineage-the development of a (plant) embryo from a single ovum cell. Based on a confocal microscopy dataset, traditionally biologists manually constructed the cell lineage, starting from this observation and reasoning backward in time to establish their inheritance. To speed up this tedious process, we make use of machine learning (ML) models trained on a database of manually established cell lineages to assist the biologist in cell assignment. Most biologists, however, are not familiar with ML, nor is it clear to them which model best predicts the embryo's development. We thus have developed a visualization system that is designed to support biologists in exploring and comparing ML models, checking the model predictions, detecting possible ML model mistakes, and deciding on the most likely embryo development. To evaluate our proposed system, we deployed our interface with six biologists in an observational study. Our results show that the visual representations of machine learning are easily understandable, and our tool, LineageD+, could potentially increase biologists' working efficiency and enhance the understanding of embryos.

EBBE-Text: Explaining Neural Networks by Exploring Text Classification Decision Boundaries

While neural networks (NN) have been successfully applied to many NLP tasks, the way they function is often difficult to interpret. In this article, we focus on binary text classification via NNs and propose a new tool, which includes a visualization of the decision boundary and the distances of data elements to this boundary. This tool increases the interpretability of NN. Our approach uses two innovative views: (1) an overview of the text representation space and (2) a local view allowing data exploration around the decision boundary for various localities of this representation space. These views are integrated into a visual platform, EBBE-Text, which also contains state-of-the-art visualizations of NN representation spaces and several kinds of information obtained from the classification process. The various views are linked through numerous interactive functionalities that enable easy exploration of texts and classification results via the various complementary views. A user study shows the effectiveness of the visual encoding and a case study illustrates the benefits of using our tool for the analysis of the classifications obtained with several recent NNs and two datasets.

How Does Attention Work in Vision Transformers? A Visual Analytics Attempt

Vision transformer (ViT) expands the success of transformer models from sequential data to images. The model decomposes an image into many smaller patches and arranges them into a sequence. Multi-head self-attentions are then applied to the sequence to learn the attention between patches. Despite many successful interpretations of transformers on sequential data, little effort has been devoted to the interpretation of ViTs, and many questions remain unanswered. For example, among the numerous attention heads, which one is more important? How strong are individual patches attending to their spatial neighbors in different heads? What attention patterns have individual heads learned? In this work, we answer these questions through a visual analytics approach. Specifically, we first identify what heads are more important in ViTs by introducing multiple pruning-based metrics. Then, we profile the spatial distribution of attention strengths between patches inside individual heads, as well as the trend of attention strengths across attention layers. Third, using an autoencoder-based learning solution, we summarize all possible attention patterns that individual heads could learn. Examining the attention strengths and patterns of the important heads, we answer why they are important. Through concrete case studies with experienced deep learning experts on multiple ViTs, we validate the effectiveness of our solution that deepens the understanding of ViTs from head importance, head attention strength, and head attention pattern.

DRAVA: Aligning Human Concepts with Machine Learning Latent Dimensions for the Visual Exploration of Small Multiples

Latent vectors extracted by machine learning (ML) are widely used in data exploration (e.g., t-SNE) but suffer from a lack of interpretability. While previous studies employed disentangled representation learning (DRL) to enable more interpretable exploration, they often overlooked the potential mismatches between the concepts of humans and the semantic dimensions learned by DRL. To address this issue, we propose Drava, a visual analytics system that supports users in 1) relating the concepts of humans with the semantic dimensions of DRL and identifying mismatches, 2) providing feedback to minimize the mismatches, and 3) obtaining data insights from concept-driven exploration. Drava provides a set of visualizations and interactions based on visual piles to help users understand and refine concepts and conduct concept-driven exploration. Meanwhile, Drava employs a concept adaptor model to fine-tune the semantic dimensions of DRL based on user refinement. The usefulness of Drava is demonstrated through application scenarios and experimental validation.

A Visual Analytics System for Improving Attention-based Traffic Forecasting Models

With deep learning (DL) outperforming conventional methods for different tasks, much effort has been devoted to utilizing DL in various domains. Researchers and developers in the traffic domain have also designed and improved DL models for forecasting tasks such as estimation of traffic speed and time of arrival. However, there exist many challenges in analyzing DL models due to the black-box property of DL models and complexity of traffic data (i.e., spatio-temporal dependencies). Collaborating with domain experts, we design a visual analytics system, AttnAnalyzer, that enables users to explore how DL models make predictions by allowing effective spatio-temporal dependency analysis. The system incorporates dynamic time warping (DTW) and Granger causality tests for computational spatio-temporal dependency analysis while providing map, table, line chart, and pixel views to assist user to perform dependency and model behavior analysis. For the evaluation, we present three case studies showing how AttnAnalyzer can effectively explore model behaviors and improve model performance in two different road networks. We also provide domain expert feedback.

Visual Analysis of Neural Architecture Spaces for Summarizing Design Principles

Recent advances in artificial intelligence largely benefit from better neural network architectures. These architectures are a product of a costly process of trial-and-error. To ease this process, we develop ArchExplorer, a visual analysis method for understanding a neural architecture space and summarizing design principles. The key idea behind our method is to make the architecture space explainable by exploiting structural distances between architectures. We formulate the pairwise distance calculation as solving an all-pairs shortest path problem. To improve efficiency, we decompose this problem into a set of single-source shortest path problems. The time complexity is reduced from O(kn²N) to O(knN). Architectures are hierarchically clustered according to the distances between them. A circle-packing-based architecture visualization has been developed to convey both the global relationships between clusters and local neighborhoods of the architectures in each cluster. Two case studies and a post-analysis are presented to demonstrate the effectiveness of ArchExplorer in summarizing design principles and selecting better-performing architectures.

Visual Concept Programming: A Visual Analytics Approach to Injecting Human Intelligence at Scale

Data-centric AI has emerged as a new research area to systematically engineer the data to land AI models for real-world applications. As a core method for data-centric AI, data programming helps experts inject domain knowledge into data and label data at scale using carefully designed labeling functions (e.g., heuristic rules, logistics). Though data programming has shown great success in the NLP domain, it is challenging to program image data because of a) the challenge to describe images using visual vocabulary without human annotations and b) lacking efficient tools for data programming of images. We present Visual Concept Programming, a first-of-its-kind visual analytics approach of using visual concepts to program image data at scale while requiring a few human efforts. Our approach is built upon three unique components. It first uses a self-supervised learning approach to learn visual representation at the pixel level and extract a dictionary of visual concepts from images without using any human annotations. The visual concepts serve as building blocks of labeling functions for experts to inject their domain knowledge. We then design interactive visualizations to explore and understand visual concepts and compose labeling functions with concepts without writing code. Finally, with the composed labeling functions, users can label the image data at scale and use the labeled data to refine the pixel-wise visual representation and concept quality. We evaluate the learned pixel-wise visual representation for the downstream task of semantic segmentation to show the effectiveness and usefulness of our approach. In addition, we demonstrate how our approach tackles real-world problems of image retrieval for autonomous driving.

Interactive and Visual Prompt Engineering for Ad-hoc Task Adaptation with Large Language Models

State-of-the-art neural language models can now be used to solve ad-hoc language tasks through zero-shot prompting without the need for supervised training. This approach has gained popularity in recent years, and researchers have demonstrated prompts that achieve strong accuracy on specific NLP tasks. However, finding a prompt for new tasks requires experimentation. Different prompt templates with different wording choices lead to significant accuracy differences. PromptIDE allows users to experiment with prompt variations, visualize prompt performance, and iteratively optimize prompts. We developed a workflow that allows users to first focus on model feedback using small data before moving on to a large data regime that allows empirical grounding of promising prompts using quantitative measures of the task. The tool then allows easy deployment of the newly created ad-hoc models. We demonstrate the utility of PromptIDE (demo: http://prompt.vizhub.ai) and our workflow using several real-world use cases.

Biological Sequence Classification: A Review on Data and General Methods

With the rapid development of biotechnology, the number of biological sequences has grown exponentially. The continuous expansion of biological sequence data promotes the application of machine learning in biological sequences to construct predictive models for mining biological sequence information. There are many branches of biological sequence classification research. In this review, we mainly focus on the function and modification classification of biological sequences based on machine learning. Sequence-based prediction and analysis are the basic tasks to understand the biological functions of DNA, RNA, proteins, and peptides. However, there are hundreds of classification models developed for biological sequences, and the quite varied specific methods seem dizzying at first glance. Here, we aim to establish a long-term support website (http://lab.malab.cn/~acy/BioseqData/home.html), which provides readers with detailed information on the classification method and download links to relevant datasets. We briefly introduce the steps to build an effective model framework for biological sequence data. In addition, a brief introduction to single-cell sequencing data analysis methods and applications in biology is also included. Finally, we discuss the current challenges and future perspectives of biological sequence classification research.

Visual Analytics for RNN-Based Deep Reinforcement Learning

Deep reinforcement learning (DRL) targets to train an autonomous agent to interact with a pre-defined environment and strives to achieve specific goals through deep neural networks (DNN). Recurrent neural network (RNN) based DRL has demonstrated superior performance, as RNNs can effectively capture the temporal evolution of the environment and respond with proper agent actions. However, apart from the outstanding performance, little is known about how RNNs understand the environment internally and what has been memorized over time. Revealing these details is extremely important for deep learning experts to understand and improve DRLs, which in contrast, is also challenging due to the complicated data transformations inside these models. In this article, we propose Deep Reinforcement Learning Interactive Visual Explorer (DRLIVE), a visual analytics system to effectively explore, interpret, and diagnose RNN-based DRLs. Having focused on DRL agents trained for different Atari games, DRLIVE accomplishes three tasks: game episode exploration, RNN hidden/cell state examination, and interactive model perturbation. Using the system, one can flexibly explore a DRL agent through interactive visualizations, discover interpretable RNN cells by prioritizing RNN hidden/cell states with a set of metrics, and further diagnose the DRL model by interactively perturbing its inputs. Through concrete studies with multiple deep learning experts, we validated the efficacy of DRLIVE.

Survey on Visual Analysis of Event Sequence Data

Event sequence data record series of discrete events in the time order of occurrence. They are commonly observed in a variety of applications ranging from electronic health records to network logs, with the characteristics of large-scale, high-dimensional and heterogeneous. This high complexity of event sequence data makes it difficult for analysts to manually explore and find patterns, resulting in ever-increasing needs for computational and perceptual aids from visual analytics techniques to extract and communicate insights from event sequence datasets. In this paper, we review the state-of-the-art visual analytics approaches, characterize them with our proposed design space, and categorize them based on analytical tasks and applications. From our review of relevant literature, we have also identified several remaining research challenges and future research opportunities.

Human-centered explainability for life sciences, healthcare, and medical informatics

Rapid advances in artificial intelligence (AI) and availability of biological, medical, and healthcare data have enabled the development of a wide variety of models. Significant success has been achieved in a wide range of fields, such as genomics, protein folding, disease diagnosis, imaging, and clinical tasks. Although widely used, the inherent opacity of deep AI models has brought criticism from the research field and little adoption in clinical practice. Concurrently, there has been a significant amount of research focused on making such methods more interpretable, reviewed here, but inherent critiques of such explainability in AI (XAI), its requirements, and concerns with fairness/robustness have hampered their real-world adoption. We here discuss how user-driven XAI can be made more useful for different healthcare stakeholders through the definition of three key personas-data scientists, clinical researchers, and clinicians-and present an overview of how different XAI approaches can address their needs. For illustration, we also walk through several research and clinical examples that take advantage of XAI open-source tools, including those that help enhance the explanation of the results through visualization. This perspective thus aims to provide a guidance tool for developing explainability solutions for healthcare by empowering both subject matter experts, providing them with a survey of available tools, and explainability developers, by providing examples of how such methods can influence in practice adoption of solutions.

Geo-Context Aware Study of Vision-Based Autonomous Driving Models and Spatial Video Data

Vision-based deep learning (DL) methods have made great progress in learning autonomous driving models from large-scale crowd-sourced video datasets. They are trained to predict instantaneous driving behaviors from video data captured by on-vehicle cameras. In this paper, we develop a geo-context aware visualization system for the study of Autonomous Driving Model (ADM) predictions together with large-scale ADM video data. The visual study is seamlessly integrated with the geographical environment by combining DL model performance with geospatial visualization techniques. Model performance measures can be studied together with a set of geospatial attributes over map views. Users can also discover and compare prediction behaviors of multiple DL models in both city-wide and street-level analysis, together with road images and video contents. Therefore, the system provides a new visual exploration platform for DL model designers in autonomous driving. Use cases and domain expert evaluation show the utility and effectiveness of the visualization system.

GenNI: Human-AI Collaboration for Data-Backed Text Generation

Table2Text systems generate textual output based on structured data utilizing machine learning. These systems are essential for fluent natural language interfaces in tools such as virtual assistants; however, left to generate freely these ML systems often produce misleading or unexpected outputs. GenNI (Generation Negotiation Interface) is an interactive visual system for high-level human-AI collaboration in producing descriptive text. The tool utilizes a deep learning model designed with explicit control states. These controls allow users to globally constrain model generations, without sacrificing the representation power of the deep learning models. The visual interface makes it possible for users to interact with AI systems following a Refine-Forecast paradigm to ensure that the generation system acts in a manner human users find suitable. We report multiple use cases on two experiments that improve over uncontrolled generation approaches, while at the same time providing fine-grained control. A demo and source code are available at https://genni.vizhub.ai.

Where Can We Help? A Visual Analytics Approach to Diagnosing and Improving Semantic Segmentation of Movable Objects

Semantic segmentation is a critical component in autonomous driving and has to be thoroughly evaluated due to safety concerns. Deep neural network (DNN) based semantic segmentation models are widely used in autonomous driving. However, it is challenging to evaluate DNN-based models due to their black-box-like nature, and it is even more difficult to assess model performance for crucial objects, such as lost cargos and pedestrians, in autonomous driving applications. In this work, we propose VASS, a Visual Analytics approach to diagnosing and improving the accuracy and robustness of Semantic Segmentation models, especially for critical objects moving in various driving scenes. The key component of our approach is a context-aware spatial representation learning that extracts important spatial information of objects, such as position, size, and aspect ratio, with respect to given scene contexts. Based on this spatial representation, we first use it to create visual summarization to analyze models' performance. We then use it to guide the generation of adversarial examples to evaluate models' spatial robustness and obtain actionable insights. We demonstrate the effectiveness of VASS via two case studies of lost cargo detection and pedestrian detection in autonomous driving. For both cases, we show quantitative evaluation on the improvement of models' performance with actionable insights obtained from VASS.

Towards Visual Explainable Active Learning for Zero-Shot Classification

Zero-shot classification is a promising paradigm to solve an applicable problem when the training classes and test classes are disjoint. Achieving this usually needs experts to externalize their domain knowledge by manually specifying a class-attribute matrix to define which classes have which attributes. Designing a suitable class-attribute matrix is the key to the subsequent procedure, but this design process is tedious and trial-and-error with no guidance. This paper proposes a visual explainable active learning approach with its design and implementation called semantic navigator to solve the above problems. This approach promotes human-AI teaming with four actions (ask, explain, recommend, respond) in each interaction loop. The machine asks contrastive questions to guide humans in the thinking process of attributes. A novel visualization called semantic map explains the current status of the machine. Therefore analysts can better understand why the machine misclassifies objects. Moreover, the machine recommends the labels of classes for each attribute to ease the labeling burden. Finally, humans can steer the model by modifying the labels interactively, and the machine adjusts its recommendations. The visual explainable active learning approach improves humans' efficiency of building zero-shot classification models interactively, compared with the method without guidance. We justify our results with user studies using the standard benchmarks for zero-shot classification.

VizSnippets: Compressing Visualization Bundles Into Representative Previews for Browsing Visualization Collections

Visualization collections, accessed by platforms such as Tableau Online or Power Bl, are used by millions of people to share and access diverse analytical knowledge in the form of interactive visualization bundles. Result snippets, compact previews of these bundles, are presented to users to help them identify relevant content when browsing collections. Our engagement with Tableau product teams and review of existing snippet designs on five platforms showed us that current practices fail to help people judge the relevance of bundles because they include only the title and one image. Users frequently need to undertake the time-consuming endeavour of opening a bundle within its visualization system to examine its many views and dashboards. In response, we contribute the first systematic approach to visualization snippet design. We propose a framework for snippet design that addresses eight key challenges that we identify. We present a computational pipeline to compress the visual and textual content of bundles into representative previews that is adaptive to a provided pixel budget and provides high information density with multiple images and carefully chosen keywords. We also reflect on the method of visual inspection through random sampling to gain confidence in model and parameter choices.

VisQA: X-raying Vision and Language Reasoning in Transformers

Visual Question Answering systems target answering open-ended textual questions given input images. They are a testbed for learning high-level reasoning with a primary use in HCI, for instance assistance for the visually impaired. Recent research has shown that state-of-the-art models tend to produce answers exploiting biases and shortcuts in the training data, and sometimes do not even look at the input image, instead of performing the required reasoning steps. We present VisQA, a visual analytics tool that explores this question of reasoning vs. bias exploitation. It exposes the key element of state-of-the-art neural models - attention maps in transformers. Our working hypothesis is that reasoning steps leading to model predictions are observable from attention distributions, which are particularly useful for visualization. The design process of VisQA was motivated by well-known bias examples from the fields of deep learning and vision-language reasoning and evaluated in two ways. First, as a result of a collaboration of three fields, machine learning, vision and language reasoning, and data analytics, the work lead to a better understanding of bias exploitation of neural models for VQA, which eventually resulted in an impact on its design and training through the proposition of a method for the transfer of reasoning patterns from an oracle model. Second, we also report on the design of VisQA, and a goal-oriented evaluation of VisQA targeting the analysis of a model decision process from multiple experts, providing evidence that it makes the inner workings of models accessible to users.

M2Lens: Visualizing and Explaining Multimodal Models for Sentiment Analysis

Multimodal sentiment analysis aims to recognize people's attitudes from multiple communication channels such as verbal content (i.e., text), voice, and facial expressions. It has become a vibrant and important research topic in natural language processing. Much research focuses on modeling the complex intra- and inter-modal interactions between different communication channels. However, current multimodal models with strong performance are often deep-learning-based techniques and work like black boxes. It is not clear how models utilize multimodal information for sentiment predictions. Despite recent advances in techniques for enhancing the explainability of machine learning models, they often target unimodal scenarios (e.g., images, sentences), and little research has been done on explaining multimodal models. In this paper, we present an interactive visual analytics system, M2 Lens, to visualize and explain multimodal models for sentiment analysis. M2 Lens provides explanations on intra- and inter-modal interactions at the global, subset, and local levels. Specifically, it summarizes the influence of three typical interaction types (i.e., dominance, complement, and conflict) on the model predictions. Moreover, M2 Lens identifies frequent and influential multimodal features and supports the multi-faceted exploration of model behaviors from language, acoustic, and visual modalities. Through two case studies and expert interviews, we demonstrate our system can help users gain deep insights into the multimodal models for sentiment analysis.

A Taxonomy of Property Measures to Unify Active Learning and Human-centered Approaches to Data Labeling

Strategies for selecting the next data instance to label, in service of generating labeled data for machine learning, have been considered separately in the machine learning literature on active learning and in the visual analytics literature on human-centered approaches. We propose a unified design space for instance selection strategies to support detailed and fine-grained analysis covering both of these perspectives. We identify a concise set of 15 properties, namely measureable characteristics of datasets or of machine learning models applied to them, that cover most of the strategies in these literatures. To quantify these properties, we introduce Property Measures (PM) as fine-grained building blocks that can be used to formalize instance selection strategies. In addition, we present a taxonomy of PMs to support the description, evaluation, and generation of PMs across four dimensions: machine learning (ML) Model Output , Instance Relations , Measure Functionality , and Measure Valence . We also create computational infrastructure to support qualitative visual data analysis: a visual analytics explainer for PMs built around an implementation of PMs using cascades of eight atomic functions. It supports eight analysis tasks, covering the analysis of datasets and ML models using visual comparison within and between PMs and groups of PMs, and over time during the interactive labeling process. We iteratively refined the PM taxonomy, the explainer, and the task abstraction in parallel with each other during a two-year formative process, and show evidence of their utility through a summative evaluation with the same infrastructure. This research builds a formal baseline for the better understanding of the commonalities and differences of instance selection strategies, which can serve as the stepping stone for the synthesis of novel strategies in future work.

VADAF: Visualization for Abnormal Client Detection and Analysis in Federated Learning

Federated Learning (FL) provides a powerful solution to distributed machine learning on a large corpus of decentralized data. It ensures privacy and security by performing computation on devices (which we refer to as clients) based on local data to improve the shared global model. However, the inaccessibility of the data and the invisibility of the computation make it challenging to interpret and analyze the training process, especially to distinguish potential client anomalies. Identifying these anomalies can help experts diagnose and improve FL models. For this reason, we propose a visual analytics system, VADAF, to depict the training dynamics and facilitate analyzing potential client anomalies. Specifically, we design a visualization scheme that supports massive training dynamics in the FL environment. Moreover, we introduce an anomaly detection method to detect potential client anomalies, which are further analyzed based on both the client model’s visual and objective estimation. Three case studies have demonstrated the effectiveness of our system in understanding the FL training process and supporting abnormal client detection and analysis.

A Multidisciplinary Survey and Framework for Design and Evaluation of Explainable AI Systems

The need for interpretable and accountable intelligent systems grows along with the prevalence of artificial intelligence ( AI ) applications used in everyday life. Explainable AI ( XAI ) systems are intended to self-explain the reasoning behind system decisions and predictions. Researchers from different disciplines work together to define, design, and evaluate explainable systems. However, scholars from different disciplines focus on different objectives and fairly independent topics of XAI research, which poses challenges for identifying appropriate design and evaluation methodology and consolidating knowledge across efforts. To this end, this article presents a survey and framework intended to share knowledge and experiences of XAI design and evaluation methods across multiple disciplines. Aiming to support diverse design goals and evaluation methods in XAI research, after a thorough review of XAI related papers in the fields of machine learning, visualization, and human-computer interaction, we present a categorization of XAI design goals and evaluation methods. Our categorization presents the mapping between design goals for different XAI user groups and their evaluation methods. From our findings, we develop a framework with step-by-step design guidelines paired with evaluation methods to close the iterative design and evaluation cycles in multidisciplinary XAI teams. Further, we provide summarized ready-to-use tables of evaluation methods and recommendations for different goals in XAI research.

A Quantitative Evaluation of Global, Rule-Based Explanations of Post-Hoc, Model Agnostic Methods

Understanding the inferences of data-driven, machine-learned models can be seen as a process that discloses the relationships between their input and output. These relationships consist and can be represented as a set of inference rules. However, the models usually do not explicit these rules to their end-users who, subsequently, perceive them as black-boxes and might not trust their predictions. Therefore, scholars have proposed several methods for extracting rules from data-driven machine-learned models to explain their logic. However, limited work exists on the evaluation and comparison of these methods. This study proposes a novel comparative approach to evaluate and compare the rulesets produced by five model-agnostic, post-hoc rule extractors by employing eight quantitative metrics. Eventually, the Friedman test was employed to check whether a method consistently performed better than the others, in terms of the selected metrics, and could be considered superior. Findings demonstrate that these metrics do not provide sufficient evidence to identify superior methods over the others. However, when used together, these metrics form a tool, applicable to every rule-extraction method and machine-learned models, that is, suitable to highlight the strengths and weaknesses of the rule-extractors in various applications in an objective and straightforward manner, without any human interventions. Thus, they are capable of successfully modelling distinctively aspects of explainability, providing to researchers and practitioners vital insights on what a model has learned during its training process and how it makes its predictions.

Visual analytics tool for the interpretation of hidden states in recurrent neural networks

In this paper, we introduce a visual analytics approach aimed at helping machine learning experts analyze the hidden states of layers in recurrent neural networks. Our technique allows the user to interactively inspect how hidden states store and process information throughout the feeding of an input sequence into the network. The technique can help answer questions, such as which parts of the input data have a higher impact on the prediction and how the model correlates each hidden state configuration with a certain output. Our visual analytics approach comprises several components: First, our input visualization shows the input sequence and how it relates to the output (using color coding). In addition, hidden states are visualized through a nonlinear projection into a 2-D visualization space using t-distributed stochastic neighbor embedding to understand the shape of the space of the hidden states. Trajectories are also employed to show the details of the evolution of the hidden state configurations. Finally, a time-multi-class heatmap matrix visualizes the evolution of the expected predictions for multi-class classifiers, and a histogram indicates the distances between the hidden states within the original space. The different visualizations are shown simultaneously in multiple views and support brushing-and-linking to facilitate the analysis of the classifications and debugging for misclassified input sequences. To demonstrate the capability of our approach, we discuss two typical use cases for long short-term memory models applied to two widely used natural language processing datasets.

Initializing LSTM internal states via manifold learning

We present an approach, based on learning an intrinsic data manifold, for the initialization of the internal state values of long short-term memory (LSTM) recurrent neural networks, ensuring consistency with the initial observed input data. Exploiting the generalized synchronization concept, we argue that the converged, "mature" internal states constitute a function on this learned manifold. The dimension of this manifold then dictates the length of observed input time series data required for consistent initialization. We illustrate our approach through a partially observed chemical model system, where initializing the internal LSTM states in this fashion yields visibly improved performance. Finally, we show that learning this data manifold enables the transformation of partially observed dynamics into fully observed ones, facilitating alternative identification paths for nonlinear dynamical systems.

Classification of Explainable Artificial Intelligence Methods through Their Output Formats

Machine and deep learning have proven their utility to generate data-driven models with high accuracy and precision. However, their non-linear, complex structures are often difficult to interpret. Consequently, many scholars have developed a plethora of methods to explain their functioning and the logic of their inferences. This systematic review aimed to organise these methods into a hierarchical classification system that builds upon and extends existing taxonomies by adding a significant dimension—the output formats. The reviewed scientific papers were retrieved by conducting an initial search on Google Scholar with the keywords “explainable artificial intelligence”; “explainable machine learning”; and “interpretable machine learning”. A subsequent iterative search was carried out by checking the bibliography of these articles. The addition of the dimension of the explanation format makes the proposed classification system a practical tool for scholars, supporting them to select the most suitable type of explanation format for the problem at hand. Given the wide variety of challenges faced by researchers, the existing XAI methods provide several solutions to meet the requirements that differ considerably between the users, problems and application fields of artificial intelligence (AI). The task of identifying the most appropriate explanation can be daunting, thus the need for a classification system that helps with the selection of methods. This work concludes by critically identifying the limitations of the formats of explanations and by providing recommendations and possible future research directions on how to build a more generally applicable XAI method. Future work should be flexible enough to meet the many requirements posed by the widespread use of AI in several fields, and the new regulations.

Analyzing the Noise Robustness of Deep Neural Networks

Adversarial examples, generated by adding small but intentionally imperceptible perturbations to normal examples, can mislead deep neural networks (DNNs) to make incorrect predictions. Although much work has been done on both adversarial attack and defense, a fine-grained understanding of adversarial examples is still lacking. To address this issue, we present a visual analysis method to explain why adversarial examples are misclassified. The key is to compare and analyze the datapaths of both the adversarial and normal examples. A datapath is a group of critical neurons along with their connections. We formulate the datapath extraction as a subset selection problem and solve it by constructing and training a neural network. A multi-level visualization consisting of a network-level visualization of data flows, a layer-level visualization of feature maps, and a neuron-level visualization of learned features, has been designed to help investigate how datapaths of adversarial and normal examples diverge and merge in the prediction process. A quantitative evaluation and a case study were conducted to demonstrate the promise of our method to explain the misclassification of adversarial examples.

Visual Analytics of a Computer-Aided Diagnosis System for Pancreatic Lesions

Machine learning is a powerful and effective tool for medical image analysis to perform computer-aided diagnosis (CAD). Having great potential in improving the accuracy of a diagnosis, CAD systems are often analyzed in terms of the final accuracy, leading to a limited understanding of the internal decision process, impossibility to gain insights, and ultimately to skepticism from clinicians. We present a visual analytics approach to uncover the decision-making process of a CAD system for classifying pancreatic cystic lesions. This CAD algorithm consists of two distinct components: random forest (RF), which classifies a set of predefined features, including demographic features, and a convolutional neural network (CNN), which analyzes radiological (imaging) features of the lesions. We study the class probabilities generated by the RF and the semantical meaning of the features learned by the CNN. We also use an eye tracker to better understand which radiological features are particularly useful for a radiologist to make a diagnosis and to quantitatively compare with the features that lead the CNN to its final classification decision. Additionally, we evaluate the effects and benefits of supplying the CAD system with a case-based visual aid in a second-reader setting.

CNN Explainer: Learning Convolutional Neural Networks with Interactive Visualization

Deep learning's great success motivates many practitioners and students to learn about this exciting technology. However, it is often challenging for beginners to take their first step due to the complexity of understanding and applying deep learning. We present CNN Explainer, an interactive visualization tool designed for non-experts to learn and examine convolutional neural networks (CNNs), a foundational deep learning model architecture. Our tool addresses key challenges that novices face while learning about CNNs, which we identify from interviews with instructors and a survey with past students. CNN Explainer tightly integrates a model overview that summarizes a CNN's structure, and on-demand, dynamic visual explanation views that help users understand the underlying components of CNNs. Through smooth transitions across levels of abstraction, our tool enables users to inspect the interplay between low-level mathematical operations and high-level model structures. A qualitative user study shows that CNN Explainer helps users more easily understand the inner workings of CNNs, and is engaging and enjoyable to use. We also derive design lessons from our study. Developed using modern web technologies, CNN Explainer runs locally in users' web browsers without the need for installation or specialized hardware, broadening the public's education access to modern deep learning techniques.

Visual Neural Decomposition to Explain Multivariate Data Sets

Investigating relationships between variables in multi-dimensional data sets is a common task for data analysts and engineers. More specifically, it is often valuable to understand which ranges of which input variables lead to particular values of a given target variable. Unfortunately, with an increasing number of independent variables, this process may become cumbersome and time-consuming due to the many possible combinations that have to be explored. In this paper, we propose a novel approach to visualize correlations between input variables and a target output variable that scales to hundreds of variables. We developed a visual model based on neural networks that can be explored in a guided way to help analysts find and understand such correlations. First, we train a neural network to predict the target from the input variables. Then, we visualize the inner workings of the resulting model to help understand relations within the data set. We further introduce a new regularization term for the backpropagation algorithm that encourages the neural network to learn representations that are easier to interpret visually. We apply our method to artificial and real-world data sets to show its utility.

Attention Flows: Analyzing and Comparing Attention Mechanisms in Language Models

Advances in language modeling have led to the development of deep attention-based models that are performant across a wide variety of natural language processing (NLP) problems. These language models are typified by a pre-training process on large unlabeled text corpora and subsequently fine-tuned for specific tasks. Although considerable work has been devoted to understanding the attention mechanisms of pre-trained models, it is less understood how a model's attention mechanisms change when trained for a target NLP task. In this paper, we propose a visual analytics approach to understanding fine-tuning in attention-based language models. Our visualization, Attention Flows, is designed to support users in querying, tracing, and comparing attention within layers, across layers, and amongst attention heads in Transformer-based language models. To help users gain insight on how a classification decision is made, our design is centered on depicting classification-based attention at the deepest layer and how attention from prior layers flows throughout words in the input. Attention Flows supports the analysis of a single model, as well as the visual comparison between pre-trained and fine-tuned models via their similarities and differences. We use Attention Flows to study attention mechanisms in various sentence understanding tasks and highlight how attention evolves to address the nuances of solving these tasks.

A Visual Analytics Framework for Explaining and Diagnosing Transfer Learning Processes

Many statistical learning models hold an assumption that the training data and the future unlabeled data are drawn from the same distribution. However, this assumption is difficult to fulfill in real-world scenarios and creates barriers in reusing existing labels from similar application domains. Transfer Learning is intended to relax this assumption by modeling relationships between domains, and is often applied in deep learning applications to reduce the demand for labeled data and training time. Despite recent advances in exploring deep learning models with visual analytics tools, little work has explored the issue of explaining and diagnosing the knowledge transfer process between deep learning models. In this paper, we present a visual analytics framework for the multi-level exploration of the transfer learning processes when training deep neural networks. Our framework establishes a multi-aspect design to explain how the learned knowledge from the existing model is transferred into the new learning task when training deep neural networks. Based on a comprehensive requirement and task analysis, we employ descriptive visualization with performance measures and detailed inspections of model behaviors from the statistical, instance, feature, and model structure levels. We demonstrate our framework through two case studies on image classification by fine-tuning AlexNets to illustrate how analysts can utilize our framework.

DECE: Decision Explorer with Counterfactual Explanations for Machine Learning Models

With machine learning models being increasingly applied to various decision-making scenarios, people have spent growing efforts to make machine learning models more transparent and explainable. Among various explanation techniques, counterfactual explanations have the advantages of being human-friendly and actionable-a counterfactual explanation tells the user how to gain the desired prediction with minimal changes to the input. Besides, counterfactual explanations can also serve as efficient probes to the models' decisions. In this work, we exploit the potential of counterfactual explanations to understand and explore the behavior of machine learning models. We design DECE, an interactive visualization system that helps understand and explore a model's decisions on individual instances and data subsets, supporting users ranging from decision-subjects to model developers. DECE supports exploratory analysis of model decisions by combining the strengths of counterfactual explanations at instance- and subgroup-levels. We also introduce a set of interactions that enable users to customize the generation of counterfactual explanations to find more actionable ones that can suit their needs. Through three use cases and an expert interview, we demonstrate the effectiveness of DECE in supporting decision exploration tasks and instance explanations.

Revisiting the Modifiable Areal Unit Problem in Deep Traffic Prediction with Visual Analytics

Deep learning methods are being increasingly used for urban traffic prediction where spatiotemporal traffic data is aggregated into sequentially organized matrices that are then fed into convolution-based residual neural networks. However, the widely known modifiable areal unit problem within such aggregation processes can lead to perturbations in the network inputs. This issue can significantly destabilize the feature embeddings and the predictions - rendering deep networks much less useful for the experts. This paper approaches this challenge by leveraging unit visualization techniques that enable the investigation of many-to-many relationships between dynamically varied multi-scalar aggregations of urban traffic data and neural network predictions. Through regular exchanges with a domain expert, we design and develop a visual analytics solution that integrates 1) a Bivariate Map equipped with an advanced bivariate colormap to simultaneously depict input traffic and prediction errors across space, 2) a Moran's I Scatterplot that provides local indicators of spatial association analysis, and 3) a Multi-scale Attribution View that arranges non-linear dot plots in a tree layout to promote model analysis and comparison across scales. We evaluate our approach through a series of case studies involving a real-world dataset of Shenzhen taxi trips, and through interviews with domain experts. We observe that geographical scale variations have important impact on prediction performances, and interactive visual exploration of dynamically varying inputs and outputs benefit experts in the development of deep traffic prediction models.

HypoML: Visual Analysis for Hypothesis-based Evaluation of Machine Learning Models

In this paper, we present a visual analytics tool for enabling hypothesis-based evaluation of machine learning (ML) models. We describe a novel ML-testing framework that combines the traditional statistical hypothesis testing (commonly used in empirical research) with logical reasoning about the conclusions of multiple hypotheses. The framework defines a controlled configuration for testing a number of hypotheses as to whether and how some extra information about a "concept" or "feature" may benefit or hinder an ML model. Because reasoning multiple hypotheses is not always straightforward, we provide HypoML as a visual analysis tool, with which, the multi-thread testing results are first transformed to analytical results using statistical and logical inferences, and then to a visual representation for rapid observation of the conclusions and the logical flow between the testing results and hypotheses. We have applied HypoML to a number of hypothesized concepts, demonstrating the intuitive and explainable nature of the visual analysis.

Analyzing the Importance of Sensors for Mode of Transportation Classification

The broad availability of smartphones and Inertial Measurement Units in particular brings them into the focus of recent research. Inertial Measurement Unit data is used for a variety of tasks. One important task is the classification of the mode of transportation. In the first step, we present a deep-learning-based algorithm that combines long-short-term-memory (LSTM) layer and convolutional layer to classify eight different modes of transportation on the Sussex–Huawei Locomotion-Transportation (SHL) dataset. The inputs of our model are the accelerometer, gyroscope, linear acceleration, magnetometer, gravity and pressure values as well as the orientation information. In the second step, we analyze the contribution of each sensor modality to the classification score and to the different modes of transportation. For this analysis, we subtract the baseline confusion matrix from a confusion matrix of a network trained with a left-out sensor modality (difference confusion matrix) and we visualize the low-level features from the LSTM layers. This approach provides useful insights into the properties of the deep-learning algorithm and indicates the presence of redundant sensor modalities.

Pat-in-the-Loop: Declarative Knowledge for Controlling Neural Networks

The dazzling success of neural networks over natural language processing systems is imposing an urgent need to control their behavior with simpler, more direct declarative rules. In this paper, we propose Pat-in-the-Loop as a model to control a specific class of syntax-oriented neural networks by adding declarative rules. In Pat-in-the-Loop, distributed tree encoders allow to exploit parse trees in neural networks, heat parse trees visualize activation of parse trees, and parse subtrees are used as declarative rules in the neural network. Hence, Pat-in-the-Loop is a model to include human control in specific natural language processing (NLP)-neural network (NN) systems that exploit syntactic information, which we will generically call Pat. A pilot study on question classification showed that declarative rules representing human knowledge, injected by Pat, can be effectively used in these neural networks to ensure correctness, relevance, and cost-effective.

Designing spontaneous behavioral switching via chaotic itinerancy

Chaotic itinerancy is a frequently observed phenomenon in high-dimensional nonlinear dynamical systems and is characterized by itinerant transitions among multiple quasi-attractors. Several studies have pointed out that high-dimensional activity in animal brains can be observed to exhibit chaotic itinerancy, which is considered to play a critical role in the spontaneous behavior generation of animals. Thus, how to design desired chaotic itinerancy is a topic of great interest, particularly for neurorobotics researchers who wish to understand and implement autonomous behavioral controls. However, it is generally difficult to gain control over high-dimensional nonlinear dynamical systems. In this study, we propose a method for implementing chaotic itinerancy reproducibly in a high-dimensional chaotic neural network. We demonstrate that our method enables us to easily design both the trajectories of quasi-attractors and the transition rules among them simply by adjusting the limited number of system parameters and by using the intrinsic high-dimensional chaos.