Machine Learning is the study of methods for programming computers to learn. Computers are applied to a wide range of tasks, and for most of these it is relatively easy for programmers to design and implement the necessary software. However, there are many tasks for which this is difficult or impossible. These can be divided into four general categories. First, there are problems for which there exist no human experts. For example, in modern automated manufacturing facilities, there is a need to predict machine failures before they occur by analyzing sensor readings. Because the machines are new, there are no human experts who can be interviewed by a programmer to provide the knowledge necessary to build a computer system. A machine learning system can study recorded data and subsequent machine failures and learn prediction rules. Second, there are problems where human experts exist, but where they are unable to explain their expertise. This is the case in many perceptual tasks, such as speech recognition, hand-writing recognition, and natural language understanding. Virtually all humans exhibit expert-level abilities on these tasks, but none of them can describe the detailed steps that they follow as they perform them. Fortunately, humans can provide machines with examples of the inputs and correct outputs for these tasks, so machine learning algorithms can learn to map the inputs to the outputs. Third, there are problems where phenomena are changing rapidly. In finance, for example, people would like to predict the future behavior of the stock market, of consumer purchases, or of exchange rates. These behaviors change frequently, so that even if a programmer could construct a good predictive computer program, it would need to be rewritten frequently. A learning program can relieve the programmer of this burden by constantly modifying and tuning a set of learned prediction rules. Fourth, there are applications that need to be customized for each computer user separately. Consider, for example, a program to filter unwanted electronic mail messages. Different users will need different filters. It is unreasonable to expect each user to program his or her own rules, and it is infeasible to provide every user with a software engineer to keep the rules up-to-date. A machine learning system can learn which mail messages the user rejects and maintain the filtering rules automatically. Machine learning addresses many of the same research questions as the fields of statistics, data mining, and psychology, but with differences of emphasis. Statistics focuses on understanding the phenomena that have generated the data, often with the goal of testing different hypotheses about those phenomena. Data mining seeks to find patterns in the data that are understandable by people. Psychological studies of human learning aspire to understand the mechanisms underlying the various learning behaviors exhibited by people (concept learning, skill acquisition, strategy change, etc.).

Machine learning

[신간의 별자리x] 우리/미술, 그리고 ‘슬픔의 박물관’

Matrix Factorization Techniques for Recommender Systems

A review on empirical mode decomposition in fault diagnosis of rotating machinery

Deep neural networks: A promising tool for fault characteristic mining and intelligent diagnosis of rotating machinery with massive data

Bearing fault diagnosis is essential for monitoring rotating machinery and equipment operating conditions. They can also identify safety hazards and avoid economic losses promptly. However, there is a shortage of tagged fault data in most factories, and it is challenging to ask them to exchange labeled fault data, especially considering the data privacy needs of customers. This article proposes a swarm learning (SL) framework that combines adversarial domain networks with convolutional neural networks (CNNs) to address this problem. The framework regards every factory as an edge-computing node and solves labeled data insufficiency and privacy protection by fusing network parameters. First, a CNN is used to compute each node, and leaders are dynamically selected to merge model parameters during the training process. Second, an adversarial domain network minimizes the feature distribution variance between nodes. Finally, an SL algorithm was used to select virtual central nodes to determine the exchange process of the model parameters among different nodes. Four datasets were used to design the experiments and demonstrate the proposed approach’s reliability. The experimental results show that the proposed framework can improve computational efficiency and reduce communication costs without relying on a central server. A final shared model can also achieve enhanced accuracy in fault diagnosis at each edge node.

A Data Privacy Protection Diagnosis Framework for Multiple Machines Vibration Signals Based on a Swarm Learning Algorithm

Investigations on generalized Hjorth's parameters for machine performance degradation assessment

Machinery suffers from deterioration no matter how high its reliability is. Maintenance is an appropriate measure to ensure machinery normal condition. So performance degradation assessment is very important for maintenance decision-making. There are two very interesting aspects when degradation assessment is performed. One is to detect early fault of machinery as early as possible. Another is to estimate remaining useful life (RUL) once early fault of machinery is detected. In this paper, wavelet lifting scheme (WLS) and hidden Markov model (HMM) are used to describe current condition of gearbox and detect early gearbox faults with a dynamic threshold. After that, another model based on final failure data is proposed to predict how much time is left before a failure occurs given the current machine condition. At last, the proposed method is validated by a set of whole life gearbox data.

On-line automatic early fault detection of rotating machinery

Slurry pumps are one of the most common machines in oil sand pumping operations to pump abrasive and erosive solids and liquids from one location to another location. The impeller of a slurry pump is prone to suffer severe wear which may cause slurry pump breakdown and result in huge economic loss. Therefore, it is necessary to construct a health indicator to monitor the health evolution of the impeller. In this paper, raw slurry pump vibration signals are reprocessed through vibration signal analysis and low-pass filtering. Then, multiple statistical features are extracted from time domain and frequency domain, respectively. It should be noted that these statistical features may be correlated and redundant. To reduce the dimensionality of these statistical features, principal component analysis is conducted on these statistical features to discover significant features, namely principal components, for tracking slurry pump health condition. Industrial slurry pump vibration signals are investigated to illustrate how the developed method works. The results show that the deteriorating trend of slurry pump impeller can be well evaluated by the developed method.

Extraction of Principal Components from Multiple Statistical Features for Slurry Pump Performance Degradation Assessment

In many recommender systems, explicit information is sparse or missing. While there is a great measure of user activities, i.e., the implicit user feedback, which implicitly reflects the preference of the user. In the literature, a number of scholars have proposed algorithms around implicit feedback, such as Bayesian Personalized Ranking(BPR). BPR is designed for implicit feedback ranking, simply it is restricted to one kind of action, e.g., purchase in Amazon or clicks a website. While, in most of the E-commerce system, user actions not only contain purchase but also click, collection and cart. Various forms of the action represent different levels of user preference. At the meanwhile, the timestamps of the actions indicate temporal aspects of the user preference. In this paper, we investigate how to combine the type of activities and the timestamp to get better recommendation results. We propose an Actions and timestamp based Bayesian Personalized Ranking model, At-BPR, for personalized ranking in the E-commerce system. The experimental results prove that, compared with BPR and the extension of BPR methods, At-BPR provides better-ranking performance.

Dong Wang

Papers

A Data Privacy Protection Diagnosis Framework for Multiple Machines Vibration Signals Based on a Swarm Learning Algorithm

Investigations on generalized Hjorth's parameters for machine performance degradation assessment

On-line automatic early fault detection of rotating machinery

Extraction of Principal Components from Multiple Statistical Features for Slurry Pump Performance Degradation Assessment

User Actions and Timestamp Based Personalized Recommendation for E-Commerce System