Article
Computer Science, Artificial Intelligence
Alexandre Heuillet, Fabien Couthouis, Natalia Diaz-Rodriguez
Summary: The study explores the development of Explainable Reinforcement Learning (XRL) and the application of XAI techniques in helping to understand the behavior and internal workings of models in reinforcement learning. The evaluation focuses on studies directly linking explainability to RL, categorizing the explanation generation into transparent algorithms and post-hoc explainability. Furthermore, it reviews prominent XAI works and their potential impact on the latest advances in RL, addressing present and future everyday problems.
KNOWLEDGE-BASED SYSTEMS
(2021)
Article
Computer Science, Artificial Intelligence
Valerio La Gatta, Vincenzo Moscato, Marco Postiglione, Giancarlo Sperli
Summary: In this paper, a novel model-agnostic Explainable AI technique named CASTLE is proposed to provide rule-based explanations based on both the local and global model's workings. The framework has been evaluated on six datasets in terms of temporal efficiency, cluster quality and model significance, showing a 6% increase in interpretability compared to another state-of-the-art technique, Anchors.
EXPERT SYSTEMS WITH APPLICATIONS
(2021)
Article
Computer Science, Artificial Intelligence
Yixing Lan, Xin Xu, Qiang Fang, Yujun Zeng, Xinwang Liu, Xianjian Zhang
Summary: This paper proposes a transfer reinforcement learning approach using auto-pruned decision trees for meta-knowledge extraction. Pre-trained policies are learned in source MDPs using RL algorithms, and meta-knowledge is extracted by re-training an auto-pruned decision tree model. In target MDPs, a hybrid policy integrating meta-knowledge and policies learned on the target MDPs is generated. Experimental results demonstrate that the proposed approach outperforms other baselines in terms of learning efficiency and interpretability.
KNOWLEDGE-BASED SYSTEMS
(2022)
Article
Computer Science, Artificial Intelligence
Mara Graziani, Lidia Dutkiewicz, Davide Calvaresi, Jose Pereira Amorim, Katerina Yordanova, Mor Vered, Rahul Nair, Pedro Henriques Abreu, Tobias Blanke, Valeria Pulignano, John O. Prior, Lode Lauwaert, Wessel Reijers, Adrien Depeursinge, Vincent Andrearczyk, Henning Muller
Summary: Since its emergence in the 1960s, Artificial Intelligence (AI) has been widely applied to various technology products and fields. Machine learning, as a major part of current AI solutions, achieves high performance on various tasks through learning from data and experience. However, the interpretability of AI models, especially deep neural networks, is often challenging. Different domains have different requirements for interpretability and tools for debugging and validating models. In this paper, the authors propose a unified terminology and definition of interpretability in AI systems, aiming to improve clarity and efficiency in the regulation of ethical and reliable AI development, and to facilitate communication across interdisciplinary areas of AI.
ARTIFICIAL INTELLIGENCE REVIEW
(2023)
Article
Biology
Michael Chary, Ed W. Boyer, Michele M. Burns
Summary: This study constructed a probabilistic logic network to model how toxicologists recognize toxidromes, comparing its performance with medical toxicologists and a decision tree model. The software, dubbed Tak, performed comparably to humans on straightforward and intermediate difficulty cases, but was outperformed by humans on challenging clinical cases. The results suggest that probabilistic logic networks can perform medical reasoning comparably to humans in a restricted domain.
COMPUTERS IN BIOLOGY AND MEDICINE
(2021)
Article
Mathematics
Catalina Lozano-Murcia, Francisco P. Romero, Jesus Serrano-Guerrero, Jose A. Olivas
Summary: Machine learning is a subfield of artificial intelligence that focuses on creating algorithms capable of learning from data and making predictions. However, in actuarial science, the interpretability of these models often poses challenges, leading to concerns about their accuracy and reliability. Explainable artificial intelligence (XAI) has emerged as a solution to address these issues by facilitating the development of accurate and comprehensible models.
Article
Multidisciplinary Sciences
Thomas McGrath, Andrei Kapishnikov, Nenad Tomasev, Adam Pearce, Martin Wattenberg, Demis Hassabis, Been Kim, Ulrich Paquet, Vladimir Kramnik
Summary: AlphaZero, a neural network engine that learns chess solely by playing against itself, acquires knowledge that enables it to outperform human chess players. Despite training without access to human games or guidance, it appears to learn concepts similar to those used by human chess players, as evidenced by linear probes and behavioral analysis.
PROCEEDINGS OF THE NATIONAL ACADEMY OF SCIENCES OF THE UNITED STATES OF AMERICA
(2022)
Article
Neurosciences
Thomas Jochmann, Marc S. Seibel, Elisabeth Jochmann, Sheraz Khan, Matti S. Haemaelaeinen, Jens Haueisen
Summary: This study investigates a convolutional neural network that detects sex from clinical EEG and finds that electrocardiac artifacts leak into the classifier. However, even after removing these artifacts, the sex can still be determined from the EEG, with topographies being critical but waveforms and frequencies not important for sex detection.
HUMAN BRAIN MAPPING
(2023)
Article
Computer Science, Artificial Intelligence
Marco Virgolin, Saverio Fracaros
Summary: Counterfactual explanations (CEs) provide insights into changing algorithmic decisions. This paper examines the relationship between robustness and sparsity of CEs, and introduces definitions of robustness for sparse CEs. Experimental results show that robust CEs are more cost-effective and preferable for users.
ARTIFICIAL INTELLIGENCE
(2023)
Article
Automation & Control Systems
Alberto Barbado, Oscar Corcho
Summary: This study combines unsupervised anomaly detection techniques, domain knowledge, and interpretable machine learning models to explain abnormal fuel consumption in vehicle fleets. Results evaluated on real-world data show that this approach provides recommendations for fuel optimization adjusted to different user profiles.
ENGINEERING APPLICATIONS OF ARTIFICIAL INTELLIGENCE
(2022)
Review
Biochemistry & Molecular Biology
Ben Allen
Summary: Deep brain stimulation is a treatment that changes brain activity to control symptoms. Artificial intelligence, particularly machine learning, has been explored to improve the treatment of brain dysfunction with deep brain stimulation. Explainable artificial intelligence approaches, using models that produce interpretable solutions, have been applied to extract domain knowledge from machine learning models related to deep brain stimulation. The most common problem addressed in these studies is patient classification, followed by efforts to optimize stimulation strategies and emphasize the importance of explainable methods. The review supports the potential of artificial intelligence to personalize deep brain stimulation protocols and adapt stimulation in real time.
Article
Computer Science, Artificial Intelligence
Marco Crespi, Andrea Ferigo, Leonardo Lucio Custode, Giovanni Iacca
Summary: Multi-Agent Reinforcement Learning (MARL) has made significant progress in the past decade, but the lack of interpretability in Deep Neural Networks (DNNs) poses a challenge, especially in MARL applications. This work proposes a population-based algorithm that combines evolutionary principles with RL to train interpretable models in multi-agent systems. The proposed approach is evaluated in a highly dynamic task and demonstrates effective policies that are easy to inspect and interpret based on domain knowledge.
APPLIED SOFT COMPUTING
(2023)
Article
Materials Science, Multidisciplinary
Marc Ackermann, Deniz Iren, Yao Yao
Summary: Microstructures in steel contain information on various length scales and can be extracted from electron microscopy images. However, current approaches for selecting important microstructural features are limited, resulting in hidden features. Machine learning is used in this study to correlate important morphological features with mechanical properties, and the learned correlation is tested on different bainitic steels.
MATERIALS & DESIGN
(2023)
Article
Computer Science, Artificial Intelligence
Cristian Pachon-Garcia, Carlos Hernandez-Perez, Pedro Delicado, Veronica Vilaplana
Summary: This paper presents SurvLIMEpy, an open-source Python package that implements the SurvLIME algorithm for computing local feature importance in machine learning models for survival analysis data. The package supports various survival models and provides visualization tools to aid result interpretation. The authors conducted experiments using simulated data and open source survival datasets to demonstrate the performance of SurvLIMEpy on different models.
EXPERT SYSTEMS WITH APPLICATIONS
(2024)
Article
Computer Science, Artificial Intelligence
David Dandolo, Chiara Masiero, Mattia Carletti, Davide Dalle Pezze, Gian Antonio Susto
Summary: This paper proposes a fast interpretability approach called Accelerated Model-agnostic Explanations (AcME) for human-in-the-loop Machine Learning applications. AcME provides feature importance scores at both the global and local level and offers a what-if analysis tool to examine the impact of feature changes on model predictions. The approach achieves comparable explanation quality to state-of-the-art methods while greatly reducing computational time.
EXPERT SYSTEMS WITH APPLICATIONS
(2023)
Review
Psychiatry
Kevin P. Hill, Mark S. Gold, Charles B. Nemeroff, William McDonald, Adrienne Grzenda, Alik S. Widge, Carolyn Rodriguez, Nina Kraguljac, John H. Krystal, Linda L. Carpenter
Summary: This review aims to outline the evidence for the therapeutic use of cannabinoids for specific medical conditions and the potential side effects associated with acute and chronic cannabis use. The results show that there are currently no approved psychiatric indications for cannabinoids, and the evidence supporting their use in the treatment of psychiatric disorders is limited. The strongest evidence for cannabinoid prescription is for the management of pain and spasticity.
AMERICAN JOURNAL OF PSYCHIATRY
(2022)
Article
Neurosciences
Scott A. Wilke, Karen Lavi, Sujin Byeon, Kevin C. Donohue, Vikaas S. Sohal
Summary: Patterns of correlated prefrontal microcircuit activity are enhanced by D2R stimulation and two mechanistically distinct antidepressants (ketamine and fluoxetine). However, this D2R-driven effect is disrupted in two etiologically distinct depression models.
BIOLOGICAL PSYCHIATRY
(2022)
Article
Biochemical Research Methods
Mark J. Schatza, Ethan B. Blackwood, Sumedh S. Nagrale, Alik S. Widge
Summary: This study developed an easy to implement, fast and accurate Toolkit for Oscillatory Real-time Tracking and Estimation (TORTE) to close the loop between brain activity and behavior. By efficiently extracting oscillatory phase and amplitude and providing options for triggering closed-loop perturbations, TORTE offers a flexible and user-friendly toolkit that is compatible with various acquisition systems and experimental preparations.
JOURNAL OF NEUROSCIENCE METHODS
(2022)
Article
Chemistry, Analytical
M. A. B. S. Akhonda, Yuri Levin-Schwartz, Vince D. Calhoun, Tulay Adali
Summary: Collecting multiple related neuroimaging datasets and non-imaging data can help us understand neural and cognitive processes and predict outcomes for intervention and treatment. While methods for analyzing imaging datasets exist, there is still a lack of methods for jointly analyzing imaging and non-imaging data. This study proposes two new approaches based on independent vector analysis (IVA) to identify multivariate relationships between imaging data and behavioral features. The simulation results show better accuracy compared to current approaches, and the analysis of functional magnetic resonance imaging (fMRI) data reveals correlations between brain networks and behavioral variables.
Article
Engineering, Biomedical
Ishita Basu, Ali Yousefi, Britni Crocker, Rina Zelmann, Angelique C. Paulk, Noam Peled, Kristen K. Ellard, Daniel S. Weisholtz, G. Rees Cosgrove, Thilo Deckersbach, Uri T. Eden, Emad N. Eskandar, Darin D. Dougherty, Sydney S. Cash, Alik S. Widge
Summary: This study demonstrates that closed-loop electrical stimulation of the internal capsule can enhance cognitive control in participants with epilepsy. Decoding of task performance from electrode activity was also achieved. These findings have the potential to improve cognitive deficits in individuals with severe mental disorders.
NATURE BIOMEDICAL ENGINEERING
(2023)
Review
Psychiatry
Mara Parellada, Alvaro Andreu-Bernabeu, Monica Burdeus, Antonia San Jose Caceres, Elena Urbiola, Linda L. Carpenter, Nina V. Kraguljac, William M. McDonald, Charles B. Nemeroff, Carolyn I. Rodriguez, Alik S. Widge, Matthew W. State, Stephan J. Sanders
Summary: The aim of this study was to evaluate response biomarkers correlated with autism spectrum disorder (ASD) symptoms. A systematic review was conducted and 280 articles were included, reporting on 940 biomarkers. However, the studies showed high heterogeneity and there is currently no sufficient evidence for response biomarkers in ASD clinical trials.
AMERICAN JOURNAL OF PSYCHIATRY
(2023)
Article
Engineering, Biomedical
Dilranjan S. Wickramasuriya, Leslie J. Crofford, Alik S. Widge, Rose T. Faghih
Summary: This article introduces a new method for estimating cortisol-related energy production and sympathetic arousal based on point process and continuous-valued data, while allowing an external influence to guide the estimates. The method modifies an existing recurrent neural network (RNN) approach to enable a hybrid estimator with this capability. Experimental results demonstrate the successful estimation of energy production and sympathetic arousal, as well as the incorporation of an external influence.
IEEE TRANSACTIONS ON BIOMEDICAL ENGINEERING
(2023)
Review
Behavioral Sciences
Ryan D. Webler, Desmond J. Oathes, Sanne J. H. van Rooij, Jonathan C. Gewirtz, Ziad Nahas, Shmuel M. Lissek, Alik S. Widge
Summary: Laboratory threat extinction and exposure-based therapy both involve repeated, safe confrontation with previously threatening stimuli. However, efforts to improve exposure outcomes using rodent extinction techniques have largely failed due to differences between rodent and human neurobiology. This review proposes a comprehensive pre-clinical human research agenda to overcome these failures, using connectivity guided depolarizing brain stimulation methods and dual threat reconsolidation-extinction paradigms to map extinction relevant circuits and inform optimal integration with exposure-based therapy.
NEUROSCIENCE AND BIOBEHAVIORAL REVIEWS
(2023)
Article
Computer Science, Hardware & Architecture
Uisub Shin, Cong Ding, Virginia Woods, Alik S. Widge, Mahsa Shoaran
Summary: This letter presents a low-power SoC with neural connectivity extraction and phase-locked DBS capabilities, which can effectively regulate abnormal brain connectivity in neurological and psychiatric disorders.
IEEE SOLID-STATE CIRCUITS LETTERS
(2023)
Article
Engineering, Electrical & Electronic
Rodrigo Capobianco Guido, Tulay Adali, Emil Bjoernson, Laure Blanc-Feraud, Ulisses Braga-Neto, Behnaz Ghoraani, Christian Jutten, Alle-Jan Van der Veen, Hong Vicky Zhao, Xiaoxing Zhu
Summary: The IEEE Signal Processing Society has provided 75 years of service to the signal processing community, making significant contributions to technological advancement.
IEEE SIGNAL PROCESSING MAGAZINE
(2023)
Article
Chemistry, Analytical
Hanlu Yang, Trung Vu, Qunfang Long, Vince Calhoun, Tuelay Adali
Summary: This study proposes a framework for subgroup identification of psychiatric patients using functional connectivity profiles obtained from fMRI data. The pipeline incorporates a data-driven method and constraint-based independent component analysis to identify meaningful subgroups with similar activation patterns in certain brain areas. The identified subgroups show significant group differences in multiple meaningful brain areas.
Article
Chemistry, Analytical
Mingyu Sun, Ben Gabrielson, Mohammad Abu Baker Siddique Akhonda, Hanlu Yang, Francisco Laport, Vince Calhoun, Tuelay Adali
Summary: Joint blind source separation (JBSS) is widely used for modeling latent structures across multiple related datasets, but it is computationally prohibitive with high-dimensional data. This paper proposes a scalable JBSS method by modeling and separating the shared subspace from the data. The method achieves excellent estimation performance with significantly reduced computational costs.
Article
Materials Science, Multidisciplinary
Jacob Peloquin, Alina Kirillova, Cynthia Rudin, L. C. Brinson, Ken Gall
Summary: This research proposes a framework for quickly predicting key mechanical properties of 3D printed gyroid lattices using information about the base material and porosity of the structure. The performance of the model was then compared to numerical simulation data and demonstrated similar accuracy at a fraction of the computation time. The model development serves as an advancement in ML-driven mechanical property prediction that can be used to guide extension of current and future models.
MATERIALS & DESIGN
(2023)
Article
Multidisciplinary Sciences
Samantha M. McDonald, Emily K. Augustine, Quinn Lanners, Cynthia Rudin, L. Catherine Brinson, Matthew L. Becker
Summary: Polymers are widely used in medical products, but the diversity in commercial polymers used in medicine is surprisingly low. Machine learning has the potential to accelerate the design of polymeric biomaterials for regenerative medicine, but the lack of available and standardized characterization of medical parameters is a major obstacle.
NATURE COMMUNICATIONS
(2023)
Article
Multidisciplinary Sciences
Jacob Peloquin, Alina Kirillova, Elizabeth Mathey, Cynthia Rudin, L. Catherine Brinson, Ken Gall
Summary: This article presents a dataset of experimentally gathered tensile stress-strain curves and measured porosity values for 389 unique gyroid lattice structures manufactured using vat photopolymerization 3D printing. The data serves as a valuable resource for the development of novel printing materials and lattice geometries, and provides insights into the influence of photopolymer material properties on the printability, geometric accuracy, and mechanical performance of 3D printed lattice structures.