While much research has been devoted to learning and machine intelligence, the field is still in its infancy. In particular, a technology that will allow for heuristic exploitation of information domain regularities to reduce the time required for knowledge acquisition while concomitantly resulting in an increase in the reliability of the acquired knowledge is still lacking. Unfortunately, contemporary learning mechanisms such as neural network architectures are inherently incapable of such performance. The objective of this paper is to present a new way of looking at learning and machine intelligence which has applicability in many fields such as in robotics, intelligent agents, data fusion, and cooperative sensing. In particular, we propose to construct a new architecture, that is, a transformational architecture for learning, intelligent fusion and transference of knowledge. A System of Systems (SoS) approach is used to realize machine intelligence. Random differences are learned by the system, generalized, and made available for subsequent replay in design transformations. Cross-domain symmetries can play a major role in design generation in particular and in the design of SoSs in general. The fundamental theory of randomization is the science, which underpins the practice. This strategy is employed in the design of the Knowledge Amplification by Structural Expert Randomization or KASER system.
Interpretability of fuzzy rule-based models has always been of significant interest to the research community and the research in this area led to a number of far-reaching results. In this study, we briefly revisit the methodology and concepts of interpretability of Takagi–Sugeno (T-S) rule-based models and develop a conceptual framework involving several levels at which rules are interpreted. The layers at which interpretability is positioned are structured hierarchically by starting with the initial fuzzy set level (originating from the design of the rules), moving to information granules of finite support (where interval calculus is engaged) and finally ending up with symbols built at the higher level. As T-S rule-based models are endowed with local functions forming the conclusion parts of the rules, with the use of the principle of justifiable granularity, we develop a way of forming an interpretable conclusion in the form of information granule. To facilitate interpretability of conditions of the rules, multidimensional fuzzy sets (coming as a result of clustering) are decomposed into a Cartesian product of 1-D fuzzy sets and the quality of the resulting decomposition is evaluated. The quality of granular rules is assessed by analyzing the relationship between specificity of condition and conclusion information granules. The rules emerging at the level of symbols are further interpreted by engaging linguistic approximation, which helps approximate a collection of linguistic terms of subconditions producing a linguistic summarization in the form <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">τ</i> (inputs are <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">A</i> ) consisting of a certain linguistic quantifier <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">τ</i> . The performance of summarization is provided in the form of ranking of the relevance of the rules. Experimental studies using publicly available data are completed and analyzed.
This article presents a novel network, contribution-degree-based spiking neural network (CDSNN), which combines ideas of spiking neural network (SNN) and fuzzy set theory. In this framework, two types of information, interval and instantaneous information conveyed by the membrane potential are described by two concepts such as area under membrane potential (AUM) and firing strength. Given that the neuron with large AUM or strong firing strength would enhance the frequency of action potentials of its postsynaptic neurons, the connection between the neuron and its postsynaptic neurons should be strengthened. Combined with an idea of membership function, three contribution degrees ( <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"><tex-math notation="LaTeX">$\boldsymbol{\mu}_E$</tex-math></inline-formula> , <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"><tex-math notation="LaTeX">$\boldsymbol{\mu}_S$</tex-math></inline-formula> , and <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"><tex-math notation="LaTeX">$\boldsymbol{\mu}_{ES}$</tex-math></inline-formula> ) are defined to quantify the ability of a neuron to provide information for postsynaptic neurons. According to these three degrees, the corresponding SpikeProp learning algorithms, referred to as SPE, SPS, and SPES, are developed. Experimental results obtained on ten benchmark datasets, one high-dimensional feature dataset, one big dataset, and one time series dataset with some commonly used algorithms, networks and CDSNN demonstrate that CDSNN can achieve improved performance in terms of accuracy, generalization, precision, recall and F-measure. The article demonstrates that the mechanism by which interval-instantaneous information is simultaneously learned in a SNN is feasible.
Carbon consumption dynamic modeling is essential for energy saving, emission reduction, and green manufacturing of iron ore sintering process. This article proposes a novel adaptive weighted broad echo state learning system (AWBESLS) for carbon consumption dynamic prediction in the sintering process by integrating adaptive weights and a reservoir with echo state characteristics. Different from previous studies, the AWBESLS adaptively matches a weight to each production data to overcome the effects of anomalous data in production data and utilizes an echo state network (ESN) for catching the dynamic state in sintering process. Carbon consumption experiments using actual production data reveal the effectiveness of the AWBESLS and compare it with some state-of-the-art methods. The results show that the AWBESLS is superior to other methods in improving the prediction performance with lowest prediction error. In summary, the AWBESLS is an effective and applicable technique for dynamic modeling of the sintering process that is easily applicable for the modeling of other manufacturing processes.
As one of the most essential sources of energy, byproduct gas plays a pivotal role in the steel industry, for which the flow tendency is generally regarded as the guidance for planning and scheduling in real production. In order to obtain the numeric estimation along with its reliability, the construction of prediction intervals (PIs) is highly demanded by any practical applications as well as being long term for providing more information on future trends. Bearing this in mind, in this article, a hierarchical granular computing (HGrC)-based model is established for constructing long-term PIs, in which probabilistic modeling gives rise to a long horizon of numeric prediction, and the deployment of information granularities hierarchically extends the result to be interval-valued format. Considering that the structure of this model has a direct impact on its performance, Monte-Carlo search with a policy gradient technique is then applied for reinforcement structure learning. Compared with the existing methods, the size (length) of the granules in the proposed approach is unequal so that it becomes effective for not only periodic but also nonperiodic data. Furthermore, with the use of parallel strategy, the efficiency can be also guaranteed for real-world applications. The experimental results demonstrate that the proposed method is superior to other commonly encountered techniques, and the stability of the structure learning process behaves better when compared with other reinforcement learning approaches.
