Neural Works Predict

The development of AI based on machine learning that utilizes large amounts of data and abundant computational resources has proven that Go AI can possess the skill to defeat the strongest active players, leading to heightened expectations for its application in solving various unsolved problems. Furthermore, recent dialogue-based AI models based on large-scale language models are significantly changing the speed and quality of intellectual production. On the other hand, in the business world, digital transformation (DX) of management and business processes is being promoted to meet the diverse needs of consumers. However, in overwhelmingly many companies, the current situation is that they are either only partially introducing digital tools into their business processes or are unable to capitalize on opportunities for business transformation due to the difficulty of making investment decisions for systems based on data utilization, the lack of expertise in data analysis techniques, and the absence of personnel skilled in organizational operations that lead to results. In this seminar, we will acquire practical knowledge through demonstrations on the following topics: - Problem-solving processes based on analytical models - Handling of small data, which poses challenges in the use of analytical models, along with various case studies - Excel-based analytical model development environments effective for sharing organizational know-how - Case studies of organizational analytical model applications directly linked to solving management issues.

In the manufacturing industry, a large amount of data is handled, but to extract valuable information from the data and interpret it, various statistical analyses need to be employed. Statistical analysis, or multivariate analysis, can be used for summarizing data, confirming trends, conducting causal analysis, and making future predictions, but it is necessary to select appropriate methods according to the objectives. Regarding the recently trending topic of artificial intelligence, efficient learning requires careful consideration of how data is presented and the need to process data into a suitable form for learning. In this context, data analysis through statistical and multivariate analysis is also required. On the other hand, when trying to master statistics and multivariate analysis, traditionally, one has to learn content that is not actually used in practice through dry and uninteresting formulas. Additionally, statistical and multivariate analysis software is necessary for data analysis. However, statistical and multivariate analysis software is polarized as described below, leading to hesitation in easy implementation and utilization. In this course, we will minimize academic content and focus on practical methods of various data analyses used in manufacturing practices. Furthermore, instead of explaining dry and uninteresting formulas, we will clarify the basics and procedures of data analysis through concrete examples.

It is said that deep learning, which is a cutting-edge technology, has become a hot topic and that the artificial intelligence boom is making a comeback. While cutting-edge technology is important, it poses a challenge for engineers in the manufacturing industry to utilize it in practical development due to its high barriers to entry. Although there is a perception that artificial intelligence is difficult to apply, when focusing on the manufacturing sector, it can actually be utilized quite easily with the appropriate methods and know-how. There are mature "engineering-suitable technologies" within artificial intelligence, including deep learning, that can be safely used in manufacturing development environments. Additionally, there are many misconceptions about big data, and its original meaning is often misrepresented. By understanding the true meaning of big data and applying appropriate methods, as well as leveraging the insights of the component engineers themselves, it is possible to minimize the data required for developing high-performance artificial intelligence and to generate large amounts of data from limited datasets. In this course, we will provide a basic explanation of neural network models and MT systems, which are artificial intelligence technologies suitable for engineering, and we will explain application know-how using specific examples from the manufacturing industry.

The experimental design method allows for the analysis of phenomena involving many components with a limited number of experiments. By using this analytical method, it is possible to find the optimal combinations of components (optimal conditions for each factor) with just a few dozen experiments, even when thousands of experiments would normally be required. However, it is based on the assumption that the interaction effects of the components follow a linear model (a simple model where the effects of components are summed), which often leads to inaccurate estimations of optimal conditions in the complex interrelations found in manufacturing development. In this course, we will first explain the principles and issues of experimental design. Then, to address the shortcomings of experimental design, we will introduce a nonlinear experimental design method that combines it with deep learning (neural network models = super regression), which is a type of artificial intelligence more suitable for manufacturing development. We will provide detailed solutions for beginners considering the introduction of experimental design, those who have tried experimental design, response surface methodology, or quality engineering (Taguchi method) without success, and those who require multi-objective optimization.