While learning theory suggests different forms of explicit regularization to guarantee small generalization error, deep learning models may introduce some sort of implicit regularization that tries to find a solution with small complexity, but neither does not include a penalty term nor does not directly modifies the optimization method. Recent research in deep learning pointed to the importance of proper understanding the underlying mechanisms of implicit regularization to elucidate the nature of generalization ability. This study is aimed at looking at implicit regularization from a psychological perspective with regard to the phenomena of retrieval-induced forgetting (RIF). The findings of this study may greatly assist in solving the major problems of proper understanding the deep learning procedure, improving the generalization ability, and the capacity control.

This paper considers a problem of improving the accuracy of identifying human activity in buildings based on an ecological feature space. To solve this problem a model of logistic regression was implemented on the assumption of the unstable estimation of logistic regression parameters for near linearly separable classes. To reach a compromise between the presence of outliers and the accuracy of recognition an algorithm of anomaly detection was proposed. Computational experiments confirmed the effectiveness of the algorithm and its theoretical consistency.

Machines usually employ a guess-and-check strategy to analyze data: they take the data, make a guess, check the answer, adjust it with regard to the correct one if necessary, and try again on a new data set. An active learning environment guarantees better performance while training on less, but carefully chosen, data which reduces the costs of both annotating and analyzing large data sets. This issue becomes even more critical for deep learning applications. Human-like active learning integrates a variety of strategies and instructional models chosen by a teacher to contribute to learners’ knowledge, while machine active learning strategies lack versatile tools for shifting the focus of instruction away from knowledge transmission to learners’ knowledge construction. We approach this gap by considering an active learning environment in an educational setting. We propose a new strategy that measures the information capacity of data using the information function from the four-parameter logistic item response theory (4PL IRT). We compared the proposed strategy with the most common active learning strategies—Least Confidence and Entropy Sampling. The results of computational experiments showed that the Information Capacity strategy shares similar behavior but provides a more flexible framework for building transparent knowledge models in deep learning.