Non-incremental Learning Research Articles

Surrogate-assisted multi-objective evolutionary feature selection of generation-based fixed evolution control for time series forecasting with LSTM networks

Surrogate-assisted multi-objective evolutionary algorithms are powerful techniques to solve computationally expensive multi-objective optimization problems. In this paper, we propose a direct fitness replacement method with generation-based fixed evolution control to implement a multi-objective evolutionary algorithm that uses a surrogate model for wrapper-type feature selection, where long short-term memory is established as the learning algorithm. The importance of the work and its benefits lie in the need to reduce the excessive computational time required by conventional wrapper-type feature selection methods based on multi-objective evolutionary algorithms and LSTM networks, maintaining or improving the predictive capacity of the models. We analyze the use of incremental learning to update the surrogate model, in comparison with the conventional non-incremental learning approach. We applied these methods in real-life time series forecasting of air quality, indoor temperature in a smart building and oil temperature in electricity transformers. Multi-step ahead predictions of the forecast models obtained with different meta-learners of the surrogate model were compared by using the Diebold–Mariano statistical test on a multi-criteria performance metric. The proposed method outperformed other approaches for feature selection including, among others, methods based on surrogate-assisted multi-objective evolutionary algorithms developed by the authors in previous research, other surrogate-assisted deterministic methods for feature selection and the conventional wrapper-type feature selection method based on LSTM, improving the prediction on test dataset by 23.98%, 34.61% and 13.77%, respectively.

Policy learning from crises: lessons learned from the Italian food stamp programme

Advancing learning is a central tenet for improving public action. Recent calls for agility, robustness, prototyping and other strategies for coping with crises imply continual learning and improvement. This article contributes to challenging this ideal interpretation of the learning process. It provides conceptual and methodological tools to investigate the relationship between policy learning and policy change and sheds light on the diverse dynamics and types of learning that can emerge from crises. At the conceptual level, the article presents a learning matrix that classifies the possible outcomes in the relationship between learning and change. On methods, our research design includes process tracing, binary comparisons, and an innovative real-time approach to the study of learning. The article investigates three municipal case studies from the Italian food stamp programme implemented during the COVID-19 lockdowns. The repetition of the programme over a short period of time offers the opportunity to investigate inter-crisis learning, the process by which lessons from the first wave of implementation contributed to reforms in the second delivery. The coronavirus crisis magnified the acquisition of knowledge and provided radical inter-programme lessons – long-term, non-incremental learning beyond the management of the emergency. Yet, the findings also highlight how this window of opportunity for learning quickly closed and how certain lessons learned may be lost in the process of reform, hard to implement or are unlikely to be extrapolated across contexts.

Coronary Heart Disease Diagnosis Through Self-Organizing Map and Fuzzy Support Vector Machine with Incremental Updates

The trade-off between computation time and predictive accuracy is important in the design and implementation of clinical decision support systems. Machine learning techniques with incremental updates have proven its usefulness in analyzing large collection of medical datasets for diseases diagnosis. This research aims to develop a predictive method for heart disease diagnosis using machine learning techniques. To this end, the proposed method is developed by unsupervised and supervised learning techniques. In particular, this research relies on Principal Component Analysis (PCA), Self-Organizing Map, Fuzzy Support Vector Machine (Fuzzy SVM), and two imputation techniques for missing value imputation. Furthermore, we apply the incremental PCA and FSVM for incremental learning of the data to reduce the computation time of disease prediction. Our data analysis on two real-world datasets, Cleveland and Statlog, showed that the use of incremental Fuzzy SVM can significantly improve the accuracy of heart disease classification. The experimental results further revealed that the method is effective in reducing the computation time of disease diagnosis in relation to the non-incremental learning technique.

Context-driven Salt Seeking Test (Rats).

Changes in reward seeking behavior often occur through incremental learning based on the difference between what is expected and what actually happens. Behavioral flexibility of this sort requires experience with rewards as better or worse than expected. However, there are some instances in which behavior can change through non-incremental learning, which requires no further experience with an outcome. Such an example of non-incremental learning is the salt appetite phenomenon. In this case, animals such as rats will immediately seek out a highly-concentrated salt solution that was previously undesired when they are put in a novel state of sodium deprivation. Importantly, this adaptive salt-seeking behavior occurs despite the fact that the rats never tasted salt in the depleted state, and therefore never tasted it as a highly desirable reward. The following protocol is a method to investigate the neural circuitry mediating adaptive salt seeking using a conditioned place preference (CPP) procedure. The procedure is designed to provide an opportunity to discover possible dissociations between the neural circuitry mediating salt seeking and salt consumption to replenish the bodily deficit after sodium depletion. Additionally, this procedure is amenable to incorporating a number of neurobiological techniques for studying the brain basis of this behavior.

Research on time series data mining algorithm based on Bayesian node incremental decision tree

Aiming at the shortage of classic ID3 decision tree and C4.5 decision tree algorithm in ability of time series data mining, this paper increases Bayesian classification algorithm in the node of the decision tree algorithm as a preprocessing Bayesian node to form incremental learning decision tree. In incremental learning, the decision tree is more powerful in the mining of time series data, and the result is more robust and accurate. In order to verify the effectiveness of the algorithm, time series data mining simulation based on UCR time series data sets have been conducted, and the performance and efficiency of ID3 decision tree, Bayesian algorithm and incremental decision tree algorithm have been contrasted respectively in non-incremental learning and incremental learning. The experimental results show that in both the incremental and non-incremental time series data mining, the incremental decision tree algorithm based on Bayesian nodes optimization which can improve the classification accuracy. When optimizing the parameters of Bayesian nodes, the consumption of pruning time will be decreased, and the efficiency of the data mining is greatly improved.

Incremental approaches for updating reducts in dynamic covering information systems

In various real-world situations, there are actually a large number of dynamic covering information systems, and non-incremental learning technique is time consuming for updating approximations of sets in dynamic covering information systems. In this paper, we investigate incremental mechanisms of updating the second and sixth lower and upper approximations of sets in dynamic covering information systems with variations of attributes. Especially, we design effective algorithms for calculating the second and sixth lower and upper approximations of sets in dynamic covering information systems. The experimental results indicate that incremental algorithms outperform non-incremental algorithms in the presence of dynamic variation of attributes. Finally, we explore several examples to illustrate that the proposed approaches are feasible to perform knowledge reduction of dynamic covering information systems.

Modelling of River-Groundwater Interactions under Rainfall Events Based on a Modified Tank Model

A multitank model experiment is employed to simulate the river-groundwater interaction under rainfall events. These experiments involve coarse and fine materials and rainfall events of 45 and 65 mm/hr. We developed a modified tank model for estimation of the groundwater table and river levels in these experiments. Parameter training of our tank model includes two algorithms: (i) the nonincremental learning algorithm-based model can predict the pore water pressure (PWP) in a slope and river under a 65 mm/hr rainfall event (coarse material) with Nash–Sutcliffe efficiency (NSE) = 0.427 and −0.909 and (ii) the incremental learning algorithm-based model can predict the PWP in a slope and river with NSE = 0.994 and 0.995. Then, the river-groundwater interaction was reproduced by a numerical case. The results of the deterministic method of the numerical case and optimized method of the modified tank model matched well.

Aware Diffusion: A Semi-Holistic Routing Protocol for Wireless Sensor Networks

Routing is a challenging task in Wireless Sensor Networks (WSNs) due to the limitation in energy and hardware capabilities in WSN nodes. This challenge prompted researchers to develop routing protocols that satisfy WSNs needs. The main design objectives are reliable delivery, low energy consumption, and prolonging network lifetime. In WSNs, routing is based on local information among neighboring nodes. Routing decisions are made locally; each node will select the next hop without any clue about the other nodes on the path. Although a full knowledge about the network yields better routing, that is not feasible in WSNs due to memory limitation and to the high traffic needed to collect the needed data about all the nodes in the network. As an effort to try to overcome this disadvantage, we are proposing in this paper aware diffusion routing protocol. Aware diffusion follows a semi-holistic approach by collecting data about the available paths and uses these data to enforce healthier paths using machine learning. The data gathering is done by adding a new stage called data collection stage. In this stage, the protocol designer can determine which parameters to collect then use these parameters in enforcing the best path according to certain criteria. In our implementation of this paradigm, we are collecting total energy on the path, lowest energy level on the path, and hop count. Again, the data collected is designer and application specific. The collected data will be used to compare available paths using non-incremental learning, and the outcome will be preferring paths that meet the designer criteria. In our case, healthier and shorter paths are preferred, which will result in less power consumption, higher delivery rate, and longer network life since healthier and fewer nodes will be doing the work.

An incremental learning algorithm for constructing Boolean functions from positive and negative examples

This paper introduces an incremental algorithm for learning a Boolean function from examples. The functions are constructed in the disjunctive normal form (DNF) or the conjunctive normal form (CNF) and emphasis is placed in inferring functions with as few clauses as possible. This incremental algorithm can be combined with any existing algorithm that infers a Boolean function from examples. In this paper it is combined with the one clause at a time (OCAT) approach (Comput. Oper. Res. 21(2) (1994) 185) and (J. Global Optim. 5(1) (1994) 64) which is a non-incremental learning approach. An extensive computational study was undertaken to assess the performance characteristics of the new approach. As examples, we used binary vectors that represent text documents from different categories from the TIPSTER collection. The computational results indicate that the new algorithm is considerably more efficient and it derives more accurate Boolean functions. As it was anticipated, the Boolean functions (in DNF or CNF form) derived by the new algorithm are comprised by more clauses than the functions derived by the non-incremental approach. Scope and purpose There is a growing need for methods that can analyze information and infer patterns in a way that can be useful to the analyst. This is the core of data mining and knowledge discovery from databases. Such methods often infer a Boolean function from observations that belong to different classes. This paper presents a methodology that infers a Boolean function in an incremental setting. The proposed approach can be used in conjunction with existing algorithms that infer a Boolean function from examples. The paper presents both algorithmic developments and also an extensive empirical study. The results presented in the paper suggest that the proposed incremental approach is highly promising.

A note on the utility of incremental learning

Historically, inductive machine learning has focused on nonincremental learning tasks, i.e., where the training set can be constructed a priori and learning stops once this set has been duly proces...

INCREMENTAL LEARNING OF PRODUCTION RULES FROM EXAMPLES UNDER UNCERTAINTY: A ROUGH SET APPROACH

This paper introduces an algorithm, LEM3, for incremental learning of production rules from examples. Based on the concept of rough sets introduced by Pawlak, LEM3 is capable of learning rules from consistent as well as inconsistent examples. In LEM3, rules are generated by using the rule-generating procedure implemented in a nonincremental learning program LEM2. Consequently, the rules learned by LEM3 do not use redundant attribute-value pairs. One major feature of LEM3 is the incorporation of a separate global data structure for storing information learned from new examples. The global data structure is updated on an example by example basis, and it provides all the essential information for the incremental updating of lower and upper approximations of a concept and the generating of rules. This separation of learned knowledge from rule-generating procedures provides a more modular design of learning systems.