Recent research advances on interactive machine learning

Closed-Loop Uncertainty: The Evaluation and Calibration of Uncertainty for Human-Machine Teams under Data Drift

Though an accurate measurement of entropy, or more generally uncertainty, is critical to the success of human-machine teams, the evaluation of the accuracy of such metrics as a probability of machine correctness is often aggregated and not assessed as an iterative control process. The entropy of the decisions made by human-machine teams may not be accurately measured under cold start or at times of data drift unless disagreements between the human and machine are immediately fed back to the classifier iteratively. In this study, we present a stochastic framework by which an uncertainty model may be evaluated iteratively as a probability of machine correctness. We target a novel problem, referred to as the threshold selection problem, which involves a user subjectively selecting the point at which a signal transitions to a low state. This problem is designed to be simple and replicable for human-machine experimentation while exhibiting properties of more complex applications. Finally, we explore the potential of incorporating feedback of machine correctness into a baseline naïve Bayes uncertainty model with a novel reinforcement learning approach. The approach refines a baseline uncertainty model by incorporating machine correctness at every iteration. Experiments are conducted over a large number of realizations to properly evaluate uncertainty at each iteration of the human-machine team. Results show that our novel approach, called closed-loop uncertainty, outperforms the baseline in every case, yielding about 45% improvement on average.

Trench visualisation from a semiautonomous excavator with a base grid map using a TOF 2D profilometer

Abstract

Real-time, three-dimensional (3D) visualisation technology can be used at construction sites to improve the quality of work. A 3D view of the landscape under work can be compared to a target 3D model of the landscape to conveniently show needed excavation tasks to a human excavator operator or to show the progress of an autonomous excavator. The purpose of this study was to demonstrate surface visualisation from measurements taken with a pulsed time-of-flight (TOF) 2D profilometer on-board a semiautonomous excavator. The semiautomatic excavator was implemented by recording the feedback script parameters from the work performed on the excavator by a human driver. 3D visualisation maps based on the triangle mesh technique were generated from the 3D point cloud using measurements of the trenches dug by a human and an autonomous excavator. The accuracy of the 3D maps was evaluated by comparing them to a high-resolution commercial 3D scanner. An analysis of the results shows that the 2D profilometer attached to the excavator can achieve almost the same 3D results as a high-quality on-site static commercial 3D scanner, whilst more easily providing an unobstructed view of the trench during operation (a 3D scanner placed next to a deep trench might not have a full view of the trench). The main technical advantages of our 2D profilometer are its compact size, measurement speed, lack of moving parts, robustness, low-cost technology that enables visualisations from a unique viewpoint on the boom of the excavator, and readiness for real-time control of the excavator’s system. This research is expected to encourage the efficiency of the digging process in the future, as well as to provide a remarkable view of trench work using an excavator as a moving platform to facilitate data visualisation.

Graphical abstract

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.

Explaining Semi-Supervised Text Alignment Through Visualization

The analysis of variance in complex text traditions is an arduous task when carried out manually. Text alignment algorithms provide domain experts with a robust alternative to such repetitive tasks. Existing white-box approaches allow the digital humanities to establish syntax-based metrics taking into account the spelling, morphology and order of words. However, they produce limited results, as semantic meanings are typically not taken into account. Our interdisciplinary collaboration between visualization and digital humanities combined a semi-supervised text alignment approach based on word embeddings that take not only syntactic but also semantic text features into account, thereby improving the overall quality of the alignment. In our collaboration, we developed different visual interfaces that communicate the word distribution in high-dimensional vector space generated by the underlying neural network for increased transparency, assessment of the tool's reliability and overall improved hypothesis generation. We further offer visual means to enable the expert reader to feed domain knowledge into the system at multiple levels with the aim of improving both the product and the process of text alignment. This ultimately illustrates how visualization can engage with and augment complex modes of reading in the humanities.

Constructing Explainable Classifiers from the Start—Enabling Human-in-the Loop Machine Learning

Interactive machine learning (IML) enables the incorporation of human expertise because the human participates in the construction of the learned model. Moreover, with human-in-the-loop machine learning (HITL-ML), the human experts drive the learning, and they can steer the learning objective not only for accuracy but perhaps for characterisation and discrimination rules, where separating one class from others is the primary objective. Moreover, this interaction enables humans to explore and gain insights into the dataset as well as validate the learned models. Validation requires transparency and interpretable classifiers. The huge relevance of understandable classification has been recently emphasised for many applications under the banner of explainable artificial intelligence (XAI). We use parallel coordinates to deploy an IML system that enables the visualisation of decision tree classifiers but also the generation of interpretable splits beyond parallel axis splits. Moreover, we show that characterisation and discrimination rules are also well communicated using parallel coordinates. In particular, we report results from the largest usability study of a IML system, confirming the merits of our approach.

Task-Based Visual Interactive Modeling: Decision Trees and Rule-Based Classifiers

Visual analytics enables the coupling of machine learning models and humans in a tightly integrated workflow, addressing various analysis tasks. Each task poses distinct demands to analysts and decision-makers. In this survey, we focus on one canonical technique for rule-based classification, namely decision tree classifiers. We provide an overview of available visualizations for decision trees with a focus on how visualizations differ with respect to 16 tasks. Further, we investigate the types of visual designs employed, and the quality measures presented. We find that (i) interactive visual analytics systems for classifier development offer a variety of visual designs, (ii) utilization tasks are sparsely covered, (iii) beyond classifier development, node-link diagrams are omnipresent, (iv) even systems designed for machine learning experts rarely feature visual representations of quality measures other than accuracy. In conclusion, we see a potential for integrating algorithmic techniques, mathematical quality measures, and tailored interactive visualizations to enable human experts to utilize their knowledge more effectively.

Human-in-the-loop machine learning: a state of the art

Researchers are defining new types of interactions between humans and machine learning algorithms generically called human-in-the-loop machine learning. Depending on who is in control of the learning process, we can identify: active learning, in which the system remains in control; interactive machine learning, in which there is a closer interaction between users and learning systems; and machine teaching, where human domain experts have control over the learning process. Aside from control, humans can also be involved in the learning process in other ways. In curriculum learning human domain experts try to impose some structure on the examples presented to improve the learning; in explainable AI the focus is on the ability of the model to explain to humans why a given solution was chosen. This collaboration between AI models and humans should not be limited only to the learning process; if we go further, we can see other terms that arise such as Usable and Useful AI. In this paper we review the state of the art of the techniques involved in the new forms of relationship between humans and ML algorithms. Our contribution is not merely listing the different approaches, but to provide definitions clarifying confusing, varied and sometimes contradictory terms; to elucidate and determine the boundaries between the different methods; and to correlate all the techniques searching for the connections and influences between them.

Active Correction for Incremental Speaker Diarization of a Collection with Human in the Loop

State of the art diarization systems now achieve decent performance but those performances are often not good enough to deploy them without any human supervision. Additionally, most approaches focus on single audio files while many use cases involving multiple recordings with recurrent speakers require the incremental processing of a collection. In this paper, we propose a framework that solicits a human in the loop to correct the clustering by answering simple questions. After defining the nature of the questions for both single file and collection of files, we propose two algorithms to list those questions and associated stopping criteria that are necessary to limit the work load on the human in the loop. Experiments performed on the ALLIES dataset show that a limited interaction with a human expert can lead to considerable improvement of up to 36.5% relative diarization error rate (DER) for single files and 33.29% for a collection.

Interactive Dimensionality Reduction for Comparative Analysis

Finding the similarities and differences between groups of datasets is a fundamental analysis task. For high-dimensional data, dimensionality reduction (DR) methods are often used to find the characteristics of each group. However, existing DR methods provide limited capability and flexibility for such comparative analysis as each method is designed only for a narrow analysis target, such as identifying factors that most differentiate groups. This paper presents an interactive DR framework where we integrate our new DR method, called ULCA (unified linear comparative analysis), with an interactive visual interface. ULCA unifies two DR schemes, discriminant analysis and contrastive learning, to support various comparative analysis tasks. To provide flexibility for comparative analysis, we develop an optimization algorithm that enables analysts to interactively refine ULCA results. Additionally, the interactive visualization interface facilitates interpretation and refinement of the ULCA results. We evaluate ULCA and the optimization algorithm to show their efficiency as well as present multiple case studies using real-world datasets to demonstrate the usefulness of this framework.

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.

Visual Analytics for Human-Centered Machine Learning

We introduce a new research area in visual analytics (VA) aiming to bridge existing gaps between methods of interactive machine learning (ML) and eXplainable Artificial Intelligence (XAI), on one side, and human minds, on the other side. The gaps are, first, a conceptual mismatch between ML/XAI outputs and human mental models and ways of reasoning, and second, a mismatch between the information quantity and level of detail and human capabilities to perceive and understand. A grand challenge is to adapt ML and XAI to human goals, concepts, values, and ways of thinking. Complementing the current efforts in XAI towards solving this challenge, VA can contribute by exploiting the potential of visualization as an effective way of communicating information to humans and a strong trigger of human abstractive perception and thinking. We propose a cross-disciplinary research framework and formulate research directions for VA.

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.

OoDAnalyzer: Interactive Analysis of Out-of-Distribution Samples

One major cause of performance degradation in predictive models is that the test samples are not well covered by the training data. Such not well-represented samples are called OoD samples. In this article, we propose OoDAnalyzer, a visual analysis approach for interactively identifying OoD samples and explaining them in context. Our approach integrates an ensemble OoD detection method and a grid-based visualization. The detection method is improved from deep ensembles by combining more features with algorithms in the same family. To better analyze and understand the OoD samples in context, we have developed a novel kNN-based grid layout algorithm motivated by Hall's theorem. The algorithm approximates the optimal layout and has O(kN²) time complexity, faster than the grid layout algorithm with overall best performance but O(N³) time complexity. Quantitative evaluation and case studies were performed on several datasets to demonstrate the effectiveness and usefulness of OoDAnalyzer.

Human social sensing is an untapped resource for computational social science

The ability to 'sense' the social environment and thereby to understand the thoughts and actions of others allows humans to fit into their social worlds, communicate and cooperate, and learn from others' experiences. Here we argue that, through the lens of computational social science, this ability can be used to advance research into human sociality. When strategically selected to represent a specific population of interest, human social sensors can help to describe and predict societal trends. In addition, their reports of how they experience their social worlds can help to build models of social dynamics that are constrained by the empirical reality of human social systems.

P6: A Declarative Language for Integrating Machine Learning in Visual Analytics

We present P6, a declarative language for building high performance visual analytics systems through its support for specifying and integrating machine learning and interactive visualization methods. As data analysis methods based on machine learning and artificial intelligence continue to advance, a visual analytics solution can leverage these methods for better exploiting large and complex data. However, integrating machine learning methods with interactive visual analysis is challenging. Existing declarative programming libraries and toolkits for visualization lack support for coupling machine learning methods. By providing a declarative language for visual analytics, P6 can empower more developers to create visual analytics applications that combine machine learning and visualization methods for data analysis and problem solving. Through a variety of example applications, we demonstrate P6's capabilities and show the benefits of using declarative specifications to build visual analytics systems. We also identify and discuss the research opportunities and challenges for declarative visual analytics.

Challenges in Evaluating Interactive Visual Machine Learning Systems

In interactive visual machine learning (IVML), humans and machine learning algorithms collaborate to achieve tasks mediated by interactive visual interfaces. This human-in-the-loop approach to machine learning brings forth not only numerous intelligibility, trust, and usability issues, but also many open questions with respect to the evaluation of the IVML system, both as separate components, and as a holistic entity that includes both human and machine intelligence. This article describes the challenges and research gaps identified in an IEEE VIS workshop on the evaluation of IVML systems.

Visual Interaction with Deep Learning Models through Collaborative Semantic Inference

Automation of tasks can have critical consequences when humans lose agency over decision processes. Deep learning models are particularly susceptible since current black-box approaches lack explainable reasoning. We argue that both the visual interface and model structure of deep learning systems need to take into account interaction design. We propose a framework of collaborative semantic inference (CSI) for the co-design of interactions and models to enable visual collaboration between humans and algorithms. The approach exposes the intermediate reasoning process of models which allows semantic interactions with the visual metaphors of a problem, which means that a user can both understand and control parts of the model reasoning process. We demonstrate the feasibility of CSI with a co-designed case study of a document summarization system.