Mehmed Kantardzic

Counterfactual explanations are increasingly vital for understanding and trusting machine learning models. This paper presents Desirability Rating-based Counterfactual (DeRaC), which is a generalized framework for generating valid counterfactual explanations applicable to classification problems with complex output spaces, including single and multi-output classification with binary and multi-class outputs. By expanding the definition of counterfactual validity through a novel “desirability rating,” the approach addresses limitations in existing methods regarding complex output spaces. The framework introduces concepts such as partially valid counterfactuals and a quantitative measure of output desirability. It can be integrated with various objective functions to identify counterfactuals that satisfy properties such as similarity, proximity, and validity. Experiments demonstrate the feasibility of systematically generating counterfactuals using existing optimization techniques, achieving varying degrees of validity and similarity; specifically, Genetic Algorithm produces consistently higher counterfactual desirability albeit at the expense of longer computation times. We observed a higher average counterfactual desirability rating of 0.871 across all tested optimization methods with Powell’s method combined with DeRaC achieving the lowest average distance of 0.897 when using a mixed-objective function. The research emphasizes the context-dependent nature of counterfactuals and lays the foundation for more transparent and trustworthy machine learning systems.

Counterfactual explanations are increasingly vital for understanding and trusting machine learning models. This study presents, Desirability Rating based Counterfactual (DeRaC), a generalized framework for generating valid counterfactual explanations applicable to multi-dimensional classification problems, including single and multi-output classification with binary and multi-label outputs. By expanding the definition of counterfactual validity through a novel “desirability rating,” the approach addresses limitations in existing methods for complex output spaces. This work details a novel framework, introducing concepts like partially valid counterfactuals and a quantitative measure of output desirability, which can be used with objective functions to find counterfactuals that also satisfy the various existing properties such as similarity, proximity, validity, actionability, etc. Experiments demonstrate the feasibility of systematically generating counterfactuals using existing optimization techniques, achieving varying degrees of validity and similarity. The research emphasizes the context-dependent nature of counterfactuals and lays the foundation for more transparent and trustworthy machine learning systems.

Deep learning offers significant advancements in the accuracy of prostate identification and classification, underscoring its potential for clinical integration. However, the opacity of deep learning models presents interpretability challenges, critical for their acceptance and utility in medical diagnosis and detection. While explanation methods have been proposed to demystify these models, enhancing their clinical viability, the efficacy and acceptance of these methods in medical tasks are not well documented. This pilot study investigates the effectiveness of deep learning explanation methods in clinical settings and identifies the attributes that radiologists consider crucial for explainability, aiming to direct future enhancements. This study reveals that while explanation methods can improve clinical task performance by up to 20%, their perceived usefulness varies, with some methods being rated poorly. Radiologists prefer explanation methods that are robust against noise, precise, and consistent. These preferences underscore the need for refining explanation methods to align with clinical expectations, emphasizing clarity, accuracy, and reliability. The findings highlight the importance of developing explanation methods that not only improve performance but also are tailored to meet the stringent requirements of clinical practice, thereby facilitating deeper trust and a broader acceptance of deep learning in medical diagnostics.

Background: The COVID-19 has introduced yet another opportunity to web-based sellers of loosely regulated substances, such as cannabidiol (CBD), to promote sales under false pretenses of curing the disease. Therefore, it has become necessary to innovate ways to identify such instances of misinformation. Objective: We sought to identify COVID-19 misinformation as it relates to the sales or promotion of CBD and used transformer-based language models to identify tweets semantically similar to quotes taken from known instances of misinformation. In this case, the known misinformation was the publicly available Warning Letters from Food and Drug Administration (FDA). Methods: We collected tweets using CBD- and COVID-19–related terms. Using a previously trained model, we extracted the tweets indicating commercialization and sales of CBD and annotated those containing COVID-19 misinformation according to the FDA definitions. We encoded the collection of tweets and misinformation quotes into sentence vectors and then calculated the cosine similarity between each quote and each tweet. This allowed us to establish a threshold to identify tweets that were making false claims regarding CBD and COVID-19 while minimizing the instances of false positives. Results: We demonstrated that by using quotes taken from Warning Letters issued by FDA to perpetrators of similar misinformation, we can identify semantically similar tweets that also contain misinformation. This was accomplished by identifying a cosine distance threshold between the sentence vectors of the Warning Letters and tweets. Conclusions: This research shows that commercial CBD or COVID-19 misinformation can potentially be identified and curbed using transformer-based language models and known prior instances of misinformation. Our approach functions without the need for labeled data, potentially reducing the time at which misinformation can be identified. Our approach shows promise in that it is easily adapted to identify other forms of misinformation related to loosely regulated substances.

BACKGROUND The COVID-19 pandemic introduced yet another medical condition for online sellers of loosely regulated substances such as cannabidiol (CBD) to falsely promote sales. As a result, it has become necessary to innovate ways to identify such instances of misinformation. OBJECTIVE We used transformer-based language models to identify COVID-19 misinformation as it relates to the sales and/or promotion of CBD, by finding tweets that are semantically similar to quotes taken from known instances of misinformation, specifically the publicly available FDA warning letters. METHODS We collected tweets using CBD and COVID-19 related terms. Using a previously trained model, we extracted the tweets indicating commercialization/sales of CBD, and annotated those containing COVID-19 misinformation, according to the FDA’s definitions. We encoded the collection of tweets and misinformation quotes into sentence vectors, and then calculated the cosine similarity between each quote and each tweet, so that a threshold could be established to identify tweets that are making false claims regarding CBD and COVID-19, while minimizing the instance of false-positives. RESULTS We demonstrated that by using quotes taken from FDA warning letters of known offenses we can identify semantically similar tweets that also contain similar misinformation. By identifying a cosine distance threshold between the sentence vector of the warning letters and the sentence vector of the tweets, we can identify tweets that contain similar forms of misinformation. CONCLUSIONS Our framework shows that commercial CBD/COVID-19 misinformation can potentially be identified and consequently curbed by using transformer-based language models and known prior instances of misinformation. Our approach functions without need for labeled data, potentially reducing the time in which misinformation could be identified. Our proposed framework shows promise in being easily adapted to identify other forms of misinformation related to loosely regulated substances, such as that related to autism, dementia, and Alzheimer’s disease.

Background: In the absence of official clinical trial information, data from social networks can be used by public health and medical researchers to assess public claims about loosely regulated substances such as cannabidiol (CBD). For example, this can be achieved by comparing the medical conditions targeted by those selling CBD against the medical conditions patients commonly treat with CBD.Objective: The objective of this study was to provide a framework for public health and medical researchers to use for identifying and analyzing the consumption and marketing of unregulated substances. Specifically, we examined CBD, which is a substance that is often presented to the public as medication despite complete evidence of efficacy and safety.Methods: We collected 567,850 tweets by searching Twitter with the Tweepy Python package using the terms “CBD” and “cannabidiol.” We trained two binary text classifiers to create two corpora of 167,755 personal use and 143,322 commercial/sales tweets. Using medical, standard, and slang dictionaries, we identified and compared the most frequently occurring medical conditions, symptoms, side effects, body parts, and other substances referenced in both corpora. In addition, to assess popular claims about the efficacy of CBD as a medical treatment circulating on Twitter, we performed sentiment analysis via the VADER (Valence Aware Dictionary for Sentiment Reasoning) model on the personal CBD tweets.Results: We found references to medically relevant terms that were unique to either personal or commercial CBD tweet classes, as well as medically relevant terms that were common to both classes. When we calculated the average sentiment scores for both personal and commercial CBD tweets referencing at least one of 17 medical conditions/symptoms terms, an overall positive sentiment was observed in both personal and commercial CBD tweets. We observed instances of negative sentiment conveyed in personal CBD tweets referencing autism, whereas CBD was also marketed multiple times as a treatment for autism within commercial tweets.Conclusions: Our proposed framework provides a tool for public health and medical researchers to analyze the consumption and marketing of unregulated substances on social networks. Our analysis showed that most users of CBD are satisfied with it in regard to the condition that it is being advertised for, with the exception of autism.

There has been increasing interest in reducing carbon footprints globally in recent years. Hence increasing share of green energy and energy efficiency are promoted by governments. Therefore, optimizing energy consumption is becoming more critical for people, companies, industries, and the environment. Predicting energy consumption more precisely means that future energy management planning can be more effective. To date, most research papers have focused on predicting residential building energy consumption; however, a large portion of the energy is consumed by industrial machines. Prediction of energy consumption of large industrial machines in real time is challenging due to concept drift, in which prediction performance deteriorates over time. In this research, a novel data-driven method multi-regime approach (MRA) was developed to better predict the energy consumption for industrial machines. Whereas most papers have focused on finding an excellent prediction model that contradicts the no-free-lunch theorem, this study concentrated on adding potential concept drift points into the prediction process. A real-world dataset was collected from a semi-autonomous grinding (SAG) mill used as a data source, and a deep neural network was utilized as a prediction model for the MRA method. The results proved that the MRA method enables the detection of multi-regimes over time and provides a highly accurate prediction performance, thanks to the dynamic model approach.

Deep learning has demonstrated remarkable accuracy analyzing images for cancer detection tasks in recent years. The accuracy that has been achieved rivals radiologists and is suitable for implementation as a clinical tool. However, a significant problem is that these models are black-box algorithms therefore they are intrinsically unexplainable. This creates a barrier for clinical implementation due to lack of trust and transparency that is a characteristic of black box algorithms. Additionally, recent regulations prevent the implementation of unexplainable models in clinical settings which further demonstrates a need for explainability. To mitigate these concerns, there have been recent studies that attempt to overcome these issues by modifying deep learning architectures or providing after-the-fact explanations. A review of the deep learning explanation literature focused on cancer detection using MR images is presented here. The gap between what clinicians deem explainable and what current methods provide is discussed and future suggestions to close this gap are provided.