Background: Conversational mental healthcare support plays a crucial role in aiding individuals with mental health concerns. Large language models (LLMs) like GPT and BERT show potential in enhancing chatbot-based therapy responses. Despite their potential, there are recognised limitations in directly deploying these LLMs for therapeutic interactions as they are trained in general context and knowledge data. The overarching aim of this study is to integrate the capabilities of both GPT and BERT with the use of specialised mental health dataset methodologies. Its goal is to enhance mental health conversations, limiting the risk and increasing quality. Methods: To achieve these aims, we will review existing chatbot methodologies from rule-based systems to advanced approaches based on cognitive behavioural therapy (CBT) principles. The study introduces a unique method which integrates a fine-tuned DialoGPT model along with the real-time capabilities of the ChatGPT 3.5 application programming interface (API). This blended combination aims to leverage the contextual awareness of LLMs and the precision of mental health-focused training. The evaluation involves a case study whereby our hybrid model is compared to traditional and standalone LLM-based chatbots. The performance is assessed using metrics such as perplexity and BLEU (Bilingual Evaluation Understudy) scores, along with subjective evaluations from end-users and mental health carers. Results: Our combined model outperforms others in conversational quality and relevance in mental healthcare. The positive feedback from patients and mental healthcare professionals is evidence of this. However, vital limitations highlight the need for further development in next-generation mental health support systems. Addressing these challenges is crucial for such technologies’ practical application and effectiveness. Conclusions: With the rise of digital mental health tools, integrating models such as LLMs transforms conversational support. The study presents a promising approach combining state-of-the-art LLMs with domain-specific fine-tuned model principles. Results suggest our combined model offers affordable and better everyday support, validated by positive feedback from patients and professionals. Our research emphasises the potential of LLMs and points towards shaping responsible and effective policies for chatbot deployment in mental healthcare. These findings will contribute to future mental healthcare chatbot development and policy guidelines, emphasising the need for balanced and effective integration of advanced models and traditional therapeutic principles.
This paper presents a novel data-driven framework for real-time State of Charge (SOC) estimation in lithium-ion battery systems using a data-driven Prognostics and Health Management (PHM) approach. The method leverages an optimized bidirectional Long Short-Term Memory (Bi-LSTM) network, trained with enhanced datasets filtered via exponentially weighted moving averages (EWMAs) and refined through SHAP-based feature attribution. Compared against a Extended Kalman Filter (EKF) and Unscented Kalman Filter (UKF) across ten diverse drive cycles, the proposed model consistently achieved superior performance, with mean absolute errors (MAEs) as low as 0.40%, outperforming EKF (0.66%) and UKF (1.36%). The Bi-LSTM model also demonstrated higher R2 values (up to 0.9999) and narrower 95% confidence intervals, confirming its precision and robustness. Real-time implementation on embedded platforms yielded inference times of 1.3–2.2 s, validating its deployability for edge applications. The framework’s model-free nature makes it adaptable to other nonlinear, time-dependent systems beyond battery SOC estimation.
14 August 2023 in Artificial Intelligence and Applications
Few-shot computer vision algorithms have enormous potential to produce promised results for innovative applications which only have a small volume of example data for training. Currently, the few-shot algorithm research focuses on applying transfer learning on deep neural networks that are pre-trained on big datasets. However, adapting the transformers requires highly cost computation resources. In addition, the overfitting or underfitting problems and low accuracy on large classes in the face validation domain are identified in our research. Thus, this paper proposed an alternative enhancement solution by adding contrasted attention to the negative face pairs and positive pairs to the training process. Extra attention is created through clustering-based face pair creation algorithms. The evaluation results show that the proposed approach sufficiently addressed the problems without requiring high-cost resources. Received: 26 April 2023 | Revised: 4 July 2023 | Accepted: 26 July 2023 Conflicts of Interest Hong Qing Yu is an Associate Editor for Artificial Intelligence and Applications, and was not involved in the editorial review or the decision to publish this article. The author declares that he has no conflicts of interest to this work. Data Availability Statement The data that support the findings of this study are openly available in Kaggle: https://www.kaggle.com/datasets/olgabelitskaya/yale-face-database and https://www.kaggle.com/datasets/jessicali9530/lfw-dataset.
In recent years, deep convolutional neural networks (DCNNs) have shown promising performance in medical image analysis, including breast lesion classification in 2D ultrasound (US) images. Despite the outstanding performance of DCNN solutions, explaining their decisions remains an open investigation. Yet, the explainability of DCNN models has become essential for healthcare systems to accept and trust the models. This paper presents a novel framework for explaining DCNN classification decisions of lesions in ultrasound images using the saliency maps linking the DCNN decisions to known cancer characteristics in the medical domain. The proposed framework consists of three main phases. First, DCNN models for classification in ultrasound images are built. Next, selected methods for visualization are applied to obtain saliency maps on the input images of the DCNN models. In the final phase, the visualization outputs and domain-known cancer characteristics are mapped. The paper then demonstrates the use of the framework for breast lesion classification from ultrasound images. We first follow the transfer learning approach and build two DCNN models. We then analyze the visualization outputs of the trained DCNN models using the EGrad-CAM and Ablation-CAM methods. We map the DCNN model decisions of benign and malignant lesions through the visualization outputs to the characteristics such as echogenicity, calcification, shape, and margin. A retrospective dataset of 1298 US images collected from different hospitals is used to evaluate the effectiveness of the framework. The test results show that these characteristics contribute differently to the benign and malignant lesions’ decisions. Our study provides the foundation for other researchers to explain the DCNN classification decisions of other cancer types.