The paper "Surveying neuro-symbolic approaches for reliable artificial intelligence of things" by Zhen Lu, Imran Afridi, Hong Jin Kang, Ivan Ruchkin, and Xi Zheng reviews the integration of Artificial Intelligence (AI) with the Internet of Things (IoT), known as AIoT. It highlights the challenges posed by the complexity and scale of AIoT, particularly in terms of testability, verifiability, and interpretability, which are addressed by the neuro-symbolic paradigm. This paradigm combines the robustness of symbolic AI with the flexibility of deep learning (DL), enabling AI systems to reason, make decisions, and generalize knowledge from large datasets more effectively. The paper identifies state-of-the-art deep learning models used in IoT, including Convolutional Neural Networks (CNNs), Recurrent Neural Networks (RNNs), Restricted Boltzmann Machines (RBMs), Autoencoders (AEs), Graph Neural Networks (GNNs), and Transformer models. It discusses the limitations of these models, such as complexity, scalability, reliance on data quality, and interpretability issues. The neuro-symbolic paradigm is introduced, detailing its evolutionary stages and current trends. It combines neural networks with symbolic AI to enhance performance and interpretability. The paper categorizes state-of-the-art neuro-symbolic approaches into six main types: knowledge graph representation learning, semantic parsing, logical reasoning, program synthesis, intelligent agents and planning, and visual question answering (VQA). Key challenges and research opportunities in enhancing AIoT reliability with neuro-symbolic approaches are also discussed, including hard-coded symbolic AI, multimodal sensor data, biased interpretability, trade-offs between interpretability and performance, and ethical and societal issues. The paper aims to provide a comprehensive review of the current state of neuro-symbolic approaches and their potential to address the limitations of traditional ML and DL models in AIoT.The paper "Surveying neuro-symbolic approaches for reliable artificial intelligence of things" by Zhen Lu, Imran Afridi, Hong Jin Kang, Ivan Ruchkin, and Xi Zheng reviews the integration of Artificial Intelligence (AI) with the Internet of Things (IoT), known as AIoT. It highlights the challenges posed by the complexity and scale of AIoT, particularly in terms of testability, verifiability, and interpretability, which are addressed by the neuro-symbolic paradigm. This paradigm combines the robustness of symbolic AI with the flexibility of deep learning (DL), enabling AI systems to reason, make decisions, and generalize knowledge from large datasets more effectively. The paper identifies state-of-the-art deep learning models used in IoT, including Convolutional Neural Networks (CNNs), Recurrent Neural Networks (RNNs), Restricted Boltzmann Machines (RBMs), Autoencoders (AEs), Graph Neural Networks (GNNs), and Transformer models. It discusses the limitations of these models, such as complexity, scalability, reliance on data quality, and interpretability issues. The neuro-symbolic paradigm is introduced, detailing its evolutionary stages and current trends. It combines neural networks with symbolic AI to enhance performance and interpretability. The paper categorizes state-of-the-art neuro-symbolic approaches into six main types: knowledge graph representation learning, semantic parsing, logical reasoning, program synthesis, intelligent agents and planning, and visual question answering (VQA). Key challenges and research opportunities in enhancing AIoT reliability with neuro-symbolic approaches are also discussed, including hard-coded symbolic AI, multimodal sensor data, biased interpretability, trade-offs between interpretability and performance, and ethical and societal issues. The paper aims to provide a comprehensive review of the current state of neuro-symbolic approaches and their potential to address the limitations of traditional ML and DL models in AIoT.