Archives of Science
Open Access

Nuclear technology; Computer learning; Intelligent robots

Introduction

Information technology has been incorporated into numerous industries over the past few decades to enhance product and service innovation and creation. Nuclear science and engineering is not renowned for being a very inventive profession, however there is growing interest in upgrading the instrumentation of both new and old nuclear reactor technologies, as well as emerging technologies like nuclear robotics (Arndt, 2015). IAEA-TECDOC-1389, 2004) and to “enhance and detect subtle variation that could remain unnoticed” (IAEA-TECDOC-1363, 2003), including the use of artificial intelligence (AI) (IAEA-TECDOC-812, 1995), in order to “address obsolescence issues, to introduce new beneficial functionality, and to improve overall performance of the plant and staff.” [1-3].

Popular machine learning methods

In the literature, there are a number of AI approaches that can be found, including “old-fashioned AI” more contemporary AI and machine learning approaches, each with distinct advantages and disadvantages. Generally speaking, the majority of machine learning techniques seek out an empirical model f that learns from a training data matrix acquired from a system, where d is the number of variables involved and n is the number of training data samples. The pattern-recognition, credit assignment, and inductive inference problems are combined in machine learning. In supervised learning, updates aim to reduce an error and enhance the algorithm’s pattern recognition capabilities by changing parameters. In unsupervised learning, updates aim to match an input. a predicted value based on the data shown. Due to their adaptability for pattern identification issues, decision trees (DT), artificial neural networks (ANNs), closest neighbour (NN), support vector machine (SVM), and naive bayes (NB) are the five most used algorithms in nuclear and radiological research applications. Fuzzy logic and evolutionary algorithms (EA) are also discussed since they have been used in some of the literature to solve nuclear and radiological-related problems as standalone algorithms or in conjunction with neural networks (i.e., neuro-fuzzy or neuroevolutionary) [4, 5].

Multi-layered ANNs have outperformed other alternative machine learning methods (such as SVM) in the new millennium when data is abundant because they have improved representation learning via numerous hidden layers and improved optimization algorithms that facilitate training. According to the universal approximation theorem, there exists a neural network with at least one hidden layer and a finite number of units that can estimate any function at any desired degree of precision.This is the case for ANNs. Additionally, the use of graphical processing units (GPUs), which are excellent at performing quick matrix and vector multiplications necessary not just for image processing but also for other tasks, led to a breakthrough in training speed [6-8].

According to the specs, GPU hardware showed a speed gain of 20 or more over central processing units (Oh and Jung, 2004), as well as higher computational scalability. (CPUs). In machine learning applications including object recognition, speech recognition, adversarial games, and controls, deep learning (DL) models have quickly advanced to become the cutting-edge technology. Convolutional neural networks and long short-term memory are the two most widely used deep learning structures for sequential information with time dependencies and object detection in photos, respectively. Although their applications in nuclear sciences have been very limited and their proper tuning requires advanced knowledge, some instances include using video to identify steel flaws underwater [9].

Conclusion

This paper provides an overview of some machine learning techniques used in nuclear science and related engineering. The goal of the authors is to enable and hasten the scientific and technological outcomes of learning-based techniques by assisting researchers in understanding the advantages of new technologies as applied to the nuclear science domain. Furthermore, it is essential that the development and application of machine learning algorithms have as their main objective the provision of quick estimation for better decision-making by users (humans in the loop), as well as the assurance of the models’ interpretability and reproducibility. Last but not least, it is recommended to leverage contemporary research accelerators that enable active (virtual) debate and partnerships in order to speed up invention. [10].

Acknowledgement

We are grateful for the support of the PWAN management, the Adamawa state ministry of education, the village heads and teachers who allowed us to conduct the study, as well as our respondents who cooperated.

Potential Conflict of Interest

No conflict or competing interests in the publication of this paper.

1. Butterfield PG, Valanis BG, Spencer PS, Lindeman CA, Nutt JG (1993) Environmental antecedents of young-onset Parkinson’s disease. Neurology 43:1150-1158.

Indexed at

2. Ward JM, Singh G, Katyal SL, Anderson LM, Kovatch RM (1985) Immunocytochemical localization of the surfactant apoprotein and Clara cell antigen in chemically induced and naturally occurring pulmonary neoplasms of mice. Am J Pathol 118(3): 493-499.

Indexed at, Google Scholar

3. Ward JM, Argilan F, Reynolds CW (1983) Immunoperoxidase localization of large granular lymphocytes in normal tissues and lesions of athymic nude rats. J Immunol 131(1): 132-139.

Indexed at, Google Scholar

4. Homburger F, Russfield AB, Weisburger JH, Lim S, Chak SP, et al.(1975) Aging changes in CD-1 HaM/ICR mice reared under standard laboratory conditions. J Natl Cancer Inst 55(1): 37-45

Indexed at, Google Scholar

5. Barlow BK, Thiruchelvam MJ, Bennice L, Cory-Slechta DA, Ballatori N, et al.(2003) Increased synaptosomal dopamine content and brain concentration of paraquat produced by selective dithiocarbamates. J Neurochem 85:1075-1086.

Indexed at

6. Behari M, Srivastava AK, Das RR, Pandey RM (2001) Risk factors of Parkinson’s disease in Indian patients. J Neurol Sci 190:49-55.

Indexed at

7. Betarbet R, Sherer TB, Di Monte DA, Greenamyre JT (2002) Mechanistic approaches to Parkinson’s disease pathogenesis. Brain Pathol 12:499-510.

Indexed at

8. Orentreich N, Matias JR, DeFelice A, Zimmerman JA (1993) Low methionine ingestion by rats extends life span. J Nutr 123(2): 269-274.

Indexed at 

9. Ooka H, Segall PE, Timiras PS (1988) Histology and survival in age-delayed low-tryptophan-fed rats. Mech Ageing Dev  43(1): 79-98.

Indexed at, Crossref

10. Segall PE, Timiras PS (1976) Patho-physiologic findings after chronic tryptophan deficiency in rats: a model for delayed growth and aging. Mech Ageing Dev 5(2): 109-124.

Indexed at, Crossref