China
Africa
Explore chapters and articles related to this topic
Agent Systems
Published in Vivek Kale, Agile Network Businesses, 2017
Agent architectures are of three types: Deliberative or symbolic architectures are those designed along the lines proposed by traditional, symbolic artificial intelligence (AI).Reactive architectures are those that eschew central symbolic representations of the agent’s environment and do not rely on symbolic reasoning.Hybrid architectures are those that try to marry the deliberative and reactive approaches.
Copy-move forgery detection of medical images using golden ball optimization
Published in International Journal of Computers and Applications, 2022
D. Suganya, K. Thirunadana Sikamani, J. Sasikala
A new method for CMFD of medical images was suggested with a view of improving the accuracy of forgery detection. The method uniformly distributed the minimum Eigenvalue-based KPs in the entire image region and evaluated the SURF features. It applied SVD for feature reduction and used GBO for optimally clustering the feature vectors. The SMGM was applied on an IDB comprising of 300 medical images, out of which 200 were tampered ones. The performances precision, specificity, sensitivity, and accuracy were evaluated and found to be superior to the existing methods. The success rate of the SMGM was found to be greater than the CSM. The SMGM can be modified to include other techniques for identifying the KPs and evaluating the features besides extending the method for block-based approaches. The developed KP-based method can be combined with block-based approach for obtaining better classification. The recent techniques of neuro-symbolic artificial intelligence and spatio-spectral image analysis can also be applied as future work to further enhance the accuracy and robustness of forgery detection methods.
The uncertainty and explainability in object recognition
Published in Journal of Experimental & Theoretical Artificial Intelligence, 2021
Humans can solve the uncertainty problem of object recognition based on a small number of samples. This is because we use higher-level knowledge about the structure of the object beyond the stimulus itself (Lake et al., 2015). This implies the necessity to develop a knowledge-representation method that can tolerate uncertainty in object recognition. Symbolic artificial intelligence subsumes uncertain knowledge-representation methods such as probabilistic logic (Lukasiewicz, 1930), fuzzy logic (Pei, 2003), and uncertainty theory (Baskiyar & Dickinson, 2003), which are derived from formal logic and symbolisation, and these can achieve abstract semantic expression. It does not help to fine-tune the topological and geometric features of objects. Moreover, these are disconnected in knowledge acquisition, presentation, and use. The most difficult semantic gap problem in artificial intelligence is the difficulty in fully describing the semantics of formal symbols. Can increasing the intermediate expression level mitigates this problem? The skeleton (Blum, 1973) tree is the basis of a formal representation of the topological and geometric features of the object for knowledge acquisition based on generalised learning. Based on the generalisation framework, a small number of similar representations are used to form a generalised explicit representation of the knowledge about the object class. Finally, the results of knowledge representation are used to conduct uncertainty reasoning experiments.
Semiotically adaptive cognition: toward the realization of remotely-operated service robots for the new normal symbiotic society
Published in Advanced Robotics, 2021
Tadahiro Taniguchi, Lotfi El Hafi, Yoshinobu Hagiwara, Akira Taniguchi, Nobutaka Shimada, Takanobu Nishiura
Hence, there is a significant scope for exploring practical machine learning methods that enable service robots to obtain semiotic knowledge of their environment in an onsite and online manner. Recent advancements in machine learning (particularly deep learning) are yet to enable service robots to gain these types of knowledge. In most cases, we still rely on rule-based or symbolic artificial intelligence (AI) tools to incorporate large parts of semiotic knowledge into service robots. This requires substantial labor resources for implementation, and thus prevents the installation of service robots. Symbolic AI systems are generally ineffective in addressing uncertainty, which is an inherent property of the real world. Furthermore, symbolic AI is deficient in learning capabilities. Therefore, a revival of the application of symbolic AI (occasionally called ‘good old-fashioned AI’) is not the solution. The deep learning community has recently conducted extensive discussions on the so-called ‘System 1/2’: System 1 represents low-level cognitive capabilities (which the deep learning approach is good at), and System 2 represents high-level cognitive capabilities (which symbolic AI is good at) [7]. Rather than a revival of the conventional approaches, the discussion over System 1/2 aims at further exploring learning-based approaches to realize high-level cognitive capabilities such as symbolic manipulation, structured knowledge representation, and logical inference in an artificial cognitive system.