Estimation of noise radiance point sources multichannel direction finding systems resolution by linear prediction method

The resolution of multichannel direction finding systems of independent noise radiance point sources is estimated quantitatively under a limited learning sample on the basis of a linear prediction methods “bank” and modified Capon algorithms.

Knowledge-aware Zero-Shot Learning: Survey and Perspective

Zero-shot learning (ZSL) which aims at predicting classes that have never appeared during the training using external knowledge (a.k.a. side information) has been widely investigated. In this paper we present a literature review towards ZSL in the perspective of external knowledge, where we categorize the external knowledge, review their methods and compare different external knowledge. With the literature review, we further discuss and outlook the role of symbolic knowledge in addressing ZSL and other machine learning sample shortage issues.

A toy model to investigate stability of AI-based dynamical systems

<p>The development of atmospheric parameterizations based on neural networks is often hampered by numerical instability issues. Previous attempts to replicate these issues in a toy model have proven ineffective. We introduce a new toy model for atmospheric dynamics, which consists in an extension of the Lorenz'63 model to a higher dimension. While neural networks trained on a single orbit can easily reproduce the dynamics of the Lorenz'63 model, they fail to reproduce the dynamics of the new toy model, leading to unstable trajectories. Instabilities become more frequent as the dimension of the new model increases, but are found to occur even in very low dimension. Training the neural network on a different learning sample, based on Latin Hypercube Sampling, solves the instability issue. Our results suggest that the design of the learning sample can significantly influence the stability of dynamical systems driven by neural networks.</p>

Parallel Algorithms for Minimal Nondeterministic Finite Automata Inference

The goal of this paper is to develop the parallel algorithms that, on input of a learning sample, identify a regular language by means of a nondeterministic finite automaton (NFA). A sample is a pair of finite sets containing positive and negative examples. Given a sample, a minimal NFA that represents the target regular language is sought. We define the task of finding an NFA, which accepts all positive examples and rejects all negative ones, as a constraint satisfaction problem, and then propose the parallel algorithms to solve the problem. The results of comprehensive computational experiments on the variety of inference tasks are reported. The question of minimizing an NFA consistent with a learning sample is computationally hard.

Electric load forecast of long-period rail transit based on fuzzy mathematics

The urban rail transit power supply system is an important part of the urban power distribution network and the power source of the rail transit system. It is responsible for providing safe and reliable electrical energy to urban rail trains and power lighting equipment. This paper processes the obtained long-period rail transit power load learning sample data matrix, according to the principle of normalization processing, effectively eliminates irregular data in the sample set and fills in possible missing data, thereby eliminating bad data or fake data for model learning. Moreover, this avoids the generation of huge errors that cause exponential growth in the model due to the increase in the learning sample size and the irregularity of the data. According to the characteristics of power load, this paper comprehensively considers the influence of temperature and date type on the maximum daily load, applies the fuzzy neural network model to the long-period load forecasting of long-period rail transit, and introduces the whole process of establishing the forecasting model in detail. Through detailed analysis of the actual data provided by the EUNITE network, the relevant factors affecting the daily maximum load were determined, and then the appropriate fuzzy input was selected to establish the corresponding fuzzy neural network prediction model, and a relatively ideal prediction result was obtained. The experimental results fully proved the great potential of fuzzy neural network in long-term power load forecasting.