We often see the term explainable in the titles of papers that describe applications based on art... more We often see the term explainable in the titles of papers that describe applications based on artificial intelligence (AI). However, the literature in explainable artificial intelligence (XAI) indicates that explanations in XAI are application-and domainspecific, hence requiring evaluation whenever they are employed to explain a model that makes decisions for a specific application problem. Additionally, the literature reveals that the performance of post-hoc methods, particularly feature attribution methods, varies substantially hinting that they do not represent a solution to AI explainability. Therefore, when using XAI methods, the quality and suitability of their information outputs should be evaluated within the specific application. For these reasons, we used a scoping review methodology to investigate papers that apply AI models and adopt methods to generate post-hoc explanations while referring to said models as explainable. This paper investigates whether the term explainable model is adopted by authors under the assumption that incorporating a post-hoc XAI method suffices to characterize a model as explainable. To inspect this problem, our review analyzes whether these papers conducted evaluations. We found that 81% of the application papers that refer to their approaches as an explainable model do not conduct any form of evaluation on the XAI method they used.
The SINDy algorithm has been successfully used to identify the governing equations of dynamical s... more The SINDy algorithm has been successfully used to identify the governing equations of dynamical systems from time series data. In this paper, we argue that this makes SINDy a potentially useful tool for causal discovery and that existing tools for causal discovery can be used to dramatically improve the performance of SINDy as tool for robust sparse modeling and system identification. We then demonstrate empirically that augmenting the SINDy algorithm with tools from causal discovery can provides engineers with a tool for learning causally robust governing equations.
This work intends to combine domain ontology with natural language processing techniques to ident... more This work intends to combine domain ontology with natural language processing techniques to identify the sentiment behind judgments aiming to provide an explanation for such polarization. Also, it intends to use the Case-Based Reasoning strategy in order to learn from past reasonings (polarizations) so they can be used in new polarizations. Some steps have been developed for treatment of negation, such as, adequacy of sentiment lexicon for a domain and adaptation of ambiguous terms classification based on past ratings. Tests were developed in two distinct areas, digital cameras and movies, to justify the model evolution until its final proposal. It was observed that the accuracy obtained by the proposed model overcomes standard statistical approaches. These results demonstrate that the model contributes to the sentiment analysis area, both as a solution that provides high levels of accuracy, as well as the possibility to present the track to achieve a particular classification.
AI methods referred to as interpretable are often discredited as inaccurate by supporters of the ... more AI methods referred to as interpretable are often discredited as inaccurate by supporters of the existence of a trade-off between interpretability and accuracy. In many problem contexts however this trade-off does not hold. This paper discusses a regression problem context to predict flight takeoff delays where the most accurate data regression model was trained via the XGBoost implementation of gradient boosted decision trees. While building an XGB-CBR Twin and converting the XGBoost feature importance into global weights in the CBR model, the resultant CBR model alone provides the most accurate local prediction, maintains the global importance to provide a global explanation of the model, and offers the most interpretable representation for local explanations. This resultant CBR model becomes a benchmark of accuracy and interpretability for this problem context, and hence it is used to evaluate the two additive feature attribute methods SHAP and LIME to explain the XGBoost regression model. The results with respect to local accuracy and feature attribution lead to potentially valuable future work.
The science and technology (S&T) cycle starts with Congress identifying areas that demand scienti... more The science and technology (S&T) cycle starts with Congress identifying areas that demand scientific advancements. The cycle continues with funding agencies distributing funds to researchers to investigate and advance these areas. These novel advancements are then reported in scholarly publications. Scientific advancements are ultimately expected to be converted into wealth, promoting economic activity, and benefiting society. It is not, however, guaranteed that the desired advances targeted by Congress are met. Scholarly publications encompass valuable knowledge about many aspects of the S&T cycle and may retain the key to many unanswered questions and support multiple, yet unoffered, services to users. Because of the data deluge, it has become necessary to rely on automated decision-making agents based on artificial intelligence (AI) methods to make decisions. This article introduces AI and proposes to relieve highly-specialized human experts from the position of decision makers and shift them to become managers of automated decision-making agents that can handle the data volume. This article will lay out research directions, technical challenges, and the benefits of applying AI in various steps of the S&T cycle.
Cyberinfrastructure refers to the computational infrastructure that supports the productivity and... more Cyberinfrastructure refers to the computational infrastructure that supports the productivity and impact of scientific fields. Many scientific fields of study rely on science gateways such as HUBzero to build cyberinfrastructure portals that offer standard features such as databases, simulation, visualization, and grid tools. This article aims to analyze requirements that align with the main goals of cyberinfrastructure. Some of the goals are to promote research, encourage interdisciplinarity, automate repetitive research steps, and to enable reproducibility and data sharing, among others, thus supporting the productivity of expenditures in science & technology. We analyze the problem cyberinfrastructure is intended to solve by focusing on entry barriers to research communities, illustrating it with the casebased reasoning community. Our study suggests additional features to existing cyberinfrastructure portals such as services to help researchers find funding and job opportunities, and where employers and funders can find researchers (e.g., expert locator engines), and the inclusion of a citation management system to facilitate creation and management of metadata for scientific workflows and other products.
Proceedings of the AAAI Conference on Artificial Intelligence
Point-of-Care Ultrasound (POCUS) refers to clinician-performed and interpreted ultrasonography at... more Point-of-Care Ultrasound (POCUS) refers to clinician-performed and interpreted ultrasonography at the patient's bedside. Interpreting these images requires a high level of expertise, which may not be available during emergencies. In this paper, we support POCUS by developing classifiers that can aid medical professionals by diagnosing whether or not a patient has pneumothorax. We decomposed the task into multiple steps, using YOLOv4 to extract relevant regions of the video and a 3D sparse coding model to represent video features. Given the difficulty in acquiring positive training videos, we trained a small-data classifier with a maximum of 15 positive and 32 negative examples. To counteract this limitation, we leveraged subject matter expert (SME) knowledge to limit the hypothesis space, thus reducing the cost of data collection. We present results using two lung ultrasound datasets and demonstrate that our model is capable of achieving performance on par with SMEs in pneumotho...
Point-of-Care Ultrasound (POCUS) refers to clinicianperformed and interpreted ultrasonography at ... more Point-of-Care Ultrasound (POCUS) refers to clinicianperformed and interpreted ultrasonography at the patient's bedside. Interpreting these images requires a high level of expertise, which may not be available during emergencies. In this paper, we support POCUS by developing classifiers that can aid medical professionals by diagnosing whether or not a patient has pneumothorax. We decomposed the task into multiple steps, using YOLOv4 to extract relevant regions of the video and a 3D sparse coding model to represent video features. Given the difficulty in acquiring positive training videos, we trained a small-data classifier with a maximum of 15 positive and 32 negative examples. To counteract this limitation, we leveraged subject matter expert (SME) knowledge to limit the hypothesis space, thus reducing the cost of data collection. We present results using two lung ultrasound datasets and demonstrate that our model is capable of achieving performance on par with SMEs in pneumothorax identification. We then developed an iOS application that runs our full system in less than 4 seconds on an iPad Pro, and less than 8 seconds on an iPhone 13 Pro, labeling key regions in the lung sonogram to provide interpretable diagnoses.
Zenodo (CERN European Organization for Nuclear Research), Aug 25, 2022
AI methods referred to as interpretable are often discredited as inaccurate by supporters of the ... more AI methods referred to as interpretable are often discredited as inaccurate by supporters of the existence of a trade-off between interpretability and accuracy. In many problem contexts however this trade-off does not hold. This paper discusses a regression problem context to predict flight takeoff delays where the most accurate data regression model was trained via the XGBoost implementation of gradient boosted decision trees. While building an XGB-CBR Twin and converting the XGBoost feature importance into global weights in the CBR model, the resultant CBR model alone provides the most accurate local prediction, maintains the global importance to provide a global explanation of the model, and offers the most interpretable representation for local explanations. This resultant CBR model becomes a benchmark of accuracy and interpretability for this problem context, and hence it is used to evaluate the two additive feature attribute methods SHAP and LIME to explain the XGBoost regression model. The results with respect to local accuracy and feature attribution lead to potentially valuable future work.
Within clinical, biomedical, and translational science, an increasing number of projects are adop... more Within clinical, biomedical, and translational science, an increasing number of projects are adopting graphs for knowledge representation. Graph‐based data models elucidate the interconnectedness among core biomedical concepts, enable data structures to be easily updated, and support intuitive queries, visualizations, and inference algorithms. However, knowledge discovery across these “knowledge graphs” (KGs) has remained difficult. Data set heterogeneity and complexity; the proliferation of ad hoc data formats; poor compliance with guidelines on findability, accessibility, interoperability, and reusability; and, in particular, the lack of a universally accepted, open‐access model for standardization across biomedical KGs has left the task of reconciling data sources to downstream consumers. Biolink Model is an open‐source data model that can be used to formalize the relationships between data structures in translational science. It incorporates object‐oriented classification and gr...
The underlying hypothesis of knowledge-based explainable artificial intelligence is: the data req... more The underlying hypothesis of knowledge-based explainable artificial intelligence is: the data required for data-centric artificial intelligence agents (e.g., neural networks) are less diverse in contents than the data required to explain the decisions of such agents to humans. The idea is that a classifier can attain high accuracy using data that express a phenomenon from one perspective whereas the audience of explanations can entail multiple stakeholders and span diverse perspectives. We hence propose to use domain knowledge to complement the data used by agents. We formulate knowledge-based explainable artificial intelligence as a supervised data classification problem aligned with the CBR methodology. In this formulation, the inputs are case problems composed of both the inputs and outputs of the data-centric agent, and case solutions, the outputs, are explanation categories obtained from domain knowledge and subject matter experts. This formulation does not typically lead to an...
Journal of Intelligent Information Systems, Aug 29, 2015
This Special Issue contains three articles providing samples of current case-based reasoning (CBR... more This Special Issue contains three articles providing samples of current case-based reasoning (CBR) research. Case-based reasoning systems perform problem-solving and interpretation based on a library of prior cases, called the case base. Given a new problem, CBR systems generate solutions by retrieving the most relevant prior case(s) in the case base, and adapting their solutions to fit new circumstances. Each new solution then becomes the basis of a new case, available to be stored. The capture of new cases can provide speedup learning by making available new cases, which require less adaptation for similar problems. In addition, the capture of new cases can increase competence if the application of a solution reveals flaws, which are then corrected to form a new case avoiding the problems. By leveraging readilyavailable prior experiences, rather than relying on extracting complex (and hard to come by) domain knowledge, case-based reasoning has enjoyed success in many application contexts and has been widely applied, giving rise to numerous fielded applications. The articles in this issue illustrate three important directions in CBR research. The primary locus of "reasoning" in CBR is the case adaptation step, but a classic problem for CBR is how to acquire the knowledge required to perform that case adaptation. One of the responses of the CBR community has been to apply machine learning methods to generating case adaptation knowledge. "Enhancing Case-Based Regression with Automatically-Generated Ensembles of Adaptations," by Jalali and Leake, shows how the ability to generate adaptation rules can be leveraged by the application of rule ensembles.
We often see the term explainable in the titles of papers that describe applications based on art... more We often see the term explainable in the titles of papers that describe applications based on artificial intelligence (AI). However, the literature in explainable artificial intelligence (XAI) indicates that explanations in XAI are application-and domainspecific, hence requiring evaluation whenever they are employed to explain a model that makes decisions for a specific application problem. Additionally, the literature reveals that the performance of post-hoc methods, particularly feature attribution methods, varies substantially hinting that they do not represent a solution to AI explainability. Therefore, when using XAI methods, the quality and suitability of their information outputs should be evaluated within the specific application. For these reasons, we used a scoping review methodology to investigate papers that apply AI models and adopt methods to generate post-hoc explanations while referring to said models as explainable. This paper investigates whether the term explainable model is adopted by authors under the assumption that incorporating a post-hoc XAI method suffices to characterize a model as explainable. To inspect this problem, our review analyzes whether these papers conducted evaluations. We found that 81% of the application papers that refer to their approaches as an explainable model do not conduct any form of evaluation on the XAI method they used.
The SINDy algorithm has been successfully used to identify the governing equations of dynamical s... more The SINDy algorithm has been successfully used to identify the governing equations of dynamical systems from time series data. In this paper, we argue that this makes SINDy a potentially useful tool for causal discovery and that existing tools for causal discovery can be used to dramatically improve the performance of SINDy as tool for robust sparse modeling and system identification. We then demonstrate empirically that augmenting the SINDy algorithm with tools from causal discovery can provides engineers with a tool for learning causally robust governing equations.
This work intends to combine domain ontology with natural language processing techniques to ident... more This work intends to combine domain ontology with natural language processing techniques to identify the sentiment behind judgments aiming to provide an explanation for such polarization. Also, it intends to use the Case-Based Reasoning strategy in order to learn from past reasonings (polarizations) so they can be used in new polarizations. Some steps have been developed for treatment of negation, such as, adequacy of sentiment lexicon for a domain and adaptation of ambiguous terms classification based on past ratings. Tests were developed in two distinct areas, digital cameras and movies, to justify the model evolution until its final proposal. It was observed that the accuracy obtained by the proposed model overcomes standard statistical approaches. These results demonstrate that the model contributes to the sentiment analysis area, both as a solution that provides high levels of accuracy, as well as the possibility to present the track to achieve a particular classification.
AI methods referred to as interpretable are often discredited as inaccurate by supporters of the ... more AI methods referred to as interpretable are often discredited as inaccurate by supporters of the existence of a trade-off between interpretability and accuracy. In many problem contexts however this trade-off does not hold. This paper discusses a regression problem context to predict flight takeoff delays where the most accurate data regression model was trained via the XGBoost implementation of gradient boosted decision trees. While building an XGB-CBR Twin and converting the XGBoost feature importance into global weights in the CBR model, the resultant CBR model alone provides the most accurate local prediction, maintains the global importance to provide a global explanation of the model, and offers the most interpretable representation for local explanations. This resultant CBR model becomes a benchmark of accuracy and interpretability for this problem context, and hence it is used to evaluate the two additive feature attribute methods SHAP and LIME to explain the XGBoost regression model. The results with respect to local accuracy and feature attribution lead to potentially valuable future work.
The science and technology (S&T) cycle starts with Congress identifying areas that demand scienti... more The science and technology (S&T) cycle starts with Congress identifying areas that demand scientific advancements. The cycle continues with funding agencies distributing funds to researchers to investigate and advance these areas. These novel advancements are then reported in scholarly publications. Scientific advancements are ultimately expected to be converted into wealth, promoting economic activity, and benefiting society. It is not, however, guaranteed that the desired advances targeted by Congress are met. Scholarly publications encompass valuable knowledge about many aspects of the S&T cycle and may retain the key to many unanswered questions and support multiple, yet unoffered, services to users. Because of the data deluge, it has become necessary to rely on automated decision-making agents based on artificial intelligence (AI) methods to make decisions. This article introduces AI and proposes to relieve highly-specialized human experts from the position of decision makers and shift them to become managers of automated decision-making agents that can handle the data volume. This article will lay out research directions, technical challenges, and the benefits of applying AI in various steps of the S&T cycle.
Cyberinfrastructure refers to the computational infrastructure that supports the productivity and... more Cyberinfrastructure refers to the computational infrastructure that supports the productivity and impact of scientific fields. Many scientific fields of study rely on science gateways such as HUBzero to build cyberinfrastructure portals that offer standard features such as databases, simulation, visualization, and grid tools. This article aims to analyze requirements that align with the main goals of cyberinfrastructure. Some of the goals are to promote research, encourage interdisciplinarity, automate repetitive research steps, and to enable reproducibility and data sharing, among others, thus supporting the productivity of expenditures in science & technology. We analyze the problem cyberinfrastructure is intended to solve by focusing on entry barriers to research communities, illustrating it with the casebased reasoning community. Our study suggests additional features to existing cyberinfrastructure portals such as services to help researchers find funding and job opportunities, and where employers and funders can find researchers (e.g., expert locator engines), and the inclusion of a citation management system to facilitate creation and management of metadata for scientific workflows and other products.
Proceedings of the AAAI Conference on Artificial Intelligence
Point-of-Care Ultrasound (POCUS) refers to clinician-performed and interpreted ultrasonography at... more Point-of-Care Ultrasound (POCUS) refers to clinician-performed and interpreted ultrasonography at the patient's bedside. Interpreting these images requires a high level of expertise, which may not be available during emergencies. In this paper, we support POCUS by developing classifiers that can aid medical professionals by diagnosing whether or not a patient has pneumothorax. We decomposed the task into multiple steps, using YOLOv4 to extract relevant regions of the video and a 3D sparse coding model to represent video features. Given the difficulty in acquiring positive training videos, we trained a small-data classifier with a maximum of 15 positive and 32 negative examples. To counteract this limitation, we leveraged subject matter expert (SME) knowledge to limit the hypothesis space, thus reducing the cost of data collection. We present results using two lung ultrasound datasets and demonstrate that our model is capable of achieving performance on par with SMEs in pneumotho...
Point-of-Care Ultrasound (POCUS) refers to clinicianperformed and interpreted ultrasonography at ... more Point-of-Care Ultrasound (POCUS) refers to clinicianperformed and interpreted ultrasonography at the patient's bedside. Interpreting these images requires a high level of expertise, which may not be available during emergencies. In this paper, we support POCUS by developing classifiers that can aid medical professionals by diagnosing whether or not a patient has pneumothorax. We decomposed the task into multiple steps, using YOLOv4 to extract relevant regions of the video and a 3D sparse coding model to represent video features. Given the difficulty in acquiring positive training videos, we trained a small-data classifier with a maximum of 15 positive and 32 negative examples. To counteract this limitation, we leveraged subject matter expert (SME) knowledge to limit the hypothesis space, thus reducing the cost of data collection. We present results using two lung ultrasound datasets and demonstrate that our model is capable of achieving performance on par with SMEs in pneumothorax identification. We then developed an iOS application that runs our full system in less than 4 seconds on an iPad Pro, and less than 8 seconds on an iPhone 13 Pro, labeling key regions in the lung sonogram to provide interpretable diagnoses.
Zenodo (CERN European Organization for Nuclear Research), Aug 25, 2022
AI methods referred to as interpretable are often discredited as inaccurate by supporters of the ... more AI methods referred to as interpretable are often discredited as inaccurate by supporters of the existence of a trade-off between interpretability and accuracy. In many problem contexts however this trade-off does not hold. This paper discusses a regression problem context to predict flight takeoff delays where the most accurate data regression model was trained via the XGBoost implementation of gradient boosted decision trees. While building an XGB-CBR Twin and converting the XGBoost feature importance into global weights in the CBR model, the resultant CBR model alone provides the most accurate local prediction, maintains the global importance to provide a global explanation of the model, and offers the most interpretable representation for local explanations. This resultant CBR model becomes a benchmark of accuracy and interpretability for this problem context, and hence it is used to evaluate the two additive feature attribute methods SHAP and LIME to explain the XGBoost regression model. The results with respect to local accuracy and feature attribution lead to potentially valuable future work.
Within clinical, biomedical, and translational science, an increasing number of projects are adop... more Within clinical, biomedical, and translational science, an increasing number of projects are adopting graphs for knowledge representation. Graph‐based data models elucidate the interconnectedness among core biomedical concepts, enable data structures to be easily updated, and support intuitive queries, visualizations, and inference algorithms. However, knowledge discovery across these “knowledge graphs” (KGs) has remained difficult. Data set heterogeneity and complexity; the proliferation of ad hoc data formats; poor compliance with guidelines on findability, accessibility, interoperability, and reusability; and, in particular, the lack of a universally accepted, open‐access model for standardization across biomedical KGs has left the task of reconciling data sources to downstream consumers. Biolink Model is an open‐source data model that can be used to formalize the relationships between data structures in translational science. It incorporates object‐oriented classification and gr...
The underlying hypothesis of knowledge-based explainable artificial intelligence is: the data req... more The underlying hypothesis of knowledge-based explainable artificial intelligence is: the data required for data-centric artificial intelligence agents (e.g., neural networks) are less diverse in contents than the data required to explain the decisions of such agents to humans. The idea is that a classifier can attain high accuracy using data that express a phenomenon from one perspective whereas the audience of explanations can entail multiple stakeholders and span diverse perspectives. We hence propose to use domain knowledge to complement the data used by agents. We formulate knowledge-based explainable artificial intelligence as a supervised data classification problem aligned with the CBR methodology. In this formulation, the inputs are case problems composed of both the inputs and outputs of the data-centric agent, and case solutions, the outputs, are explanation categories obtained from domain knowledge and subject matter experts. This formulation does not typically lead to an...
Journal of Intelligent Information Systems, Aug 29, 2015
This Special Issue contains three articles providing samples of current case-based reasoning (CBR... more This Special Issue contains three articles providing samples of current case-based reasoning (CBR) research. Case-based reasoning systems perform problem-solving and interpretation based on a library of prior cases, called the case base. Given a new problem, CBR systems generate solutions by retrieving the most relevant prior case(s) in the case base, and adapting their solutions to fit new circumstances. Each new solution then becomes the basis of a new case, available to be stored. The capture of new cases can provide speedup learning by making available new cases, which require less adaptation for similar problems. In addition, the capture of new cases can increase competence if the application of a solution reveals flaws, which are then corrected to form a new case avoiding the problems. By leveraging readilyavailable prior experiences, rather than relying on extracting complex (and hard to come by) domain knowledge, case-based reasoning has enjoyed success in many application contexts and has been widely applied, giving rise to numerous fielded applications. The articles in this issue illustrate three important directions in CBR research. The primary locus of "reasoning" in CBR is the case adaptation step, but a classic problem for CBR is how to acquire the knowledge required to perform that case adaptation. One of the responses of the CBR community has been to apply machine learning methods to generating case adaptation knowledge. "Enhancing Case-Based Regression with Automatically-Generated Ensembles of Adaptations," by Jalali and Leake, shows how the ability to generate adaptation rules can be leveraged by the application of rule ensembles.
This chapter introduces the basic concepts of Case-Based Reasoning. It does not require any previ... more This chapter introduces the basic concepts of Case-Based Reasoning. It does not require any previous knowledge about the topic. The intention is the reader should understand the principal ideas and follow the descriptions of the first examples, the remaining chapters in PART I, and PART II. This chapter provides fundamental basis for understanding the remainder of the book. In most cases no formal definitions are used. In later chapters the concepts will be extended and put on a more formal basis as well as be illustrated by examples.
Uploads
Papers by Rosina Weber