Modeling the Dynamics of Spiking Networks with Memristor-Based STDP to Solve Classification Tasks

The problem with training spiking neural networks (SNNs) is relevant due to the ultra-low power consumption these networks could exhibit when implemented in neuromorphic hardware. The ongoing progress in the fabrication of memristors, a prospective basis for analogue synapses, gives relevance to studying the possibility of SNN learning on the base of synaptic plasticity models, obtained by fitting the experimental measurements of the memristor conductance change. The dynamics of memristor conductances is (necessarily) nonlinear, because conductance changes depend on the spike timings, which neurons emit in an all-or-none fashion. The ability to solve classification tasks was previously shown for spiking network models based on the bio-inspired local learning mechanism of spike-timing-dependent plasticity (STDP), as well as with the plasticity that models the conductance change of nanocomposite (NC) memristors. Input data were presented to the network encoded into the intensities of Poisson input spike sequences. This work considers another approach for encoding input data into input spike sequences presented to the network: temporal encoding, in which an input vector is transformed into relative timing of individual input spikes. Since temporal encoding uses fewer input spikes, the processing of each input vector by the network can be faster and more energy-efficient. The aim of the current work is to show the applicability of temporal encoding to training spiking networks with three synaptic plasticity models: STDP, NC memristor approximation, and PPX memristor approximation. We assess the accuracy of the proposed approach on several benchmark classification tasks: Fisher’s Iris, Wisconsin breast cancer, and the pole balancing task (CartPole). The accuracies achieved by SNN with memristor plasticity and conventional STDP are comparable and are on par with classic machine learning approaches.

OPTIMIZATION OF THE INPUT DATA VECTOR TO IMPROVE THE NEURAL NETWORK TRAINING FOR OPC

This work explored the impact of input data structure to improve the neural network training. The impact of two variants of the input data vector on the training accuracy of the network was studied. The first version of the input vector included the intensity of the exposure radiation map. The second version of the input vector included the intensity of the exposure radiation map and IC topology.

Multi-Source Data Fusion and Target Tracking of Heterogeneous Network Based on Data Mining

Thanks to the technical development of target tracking, the multi-source data fusion and target tracking has become a hotspot in the research of huge heterogenous networks. Based on millimeter wave heterogeneous network, this paper constructs a multi-source data fusion and target tracking model. The core of the model is the data mining deep Q network (DM-DQN). Through image filling, the length of the input vector (time window) was extended from 25 to 31, with the aid of CNN heterogeneous network technology. This is to keep the length of input vector in line with that of output vector, and retain the time features of eye tracking data to the greatest extent, thereby expanding the recognition range. Experimental results show that the proposed model achieved a modified mean error of only 1.5m with a tracking time of 160s, that is, the tracking effect is ideal. That is why the DM-DQN outperformed other algorithms in total user delay. The algorithm can improve the energy efficiency of the network, while ensuring the quality of service of the user. In the first 50 iterations, DM-DQN worked poorer than structured data mining. After 50 iterations, DM-DQN began to learn the merits of the latter. After 100 iterations, both DM-DQN and structured data mining tended to be stable, and the former had the better performance. Compared with typical structured data mining, the proposed DM-DQN not only converges fast, but also boasts a relatively good performance.

Classification of Grain Storage Inventory Modes Based on Temperature Contour Map of Grain Bulk Using Back Propagation Neural Network

Inventory modes classification can reduce the workload of grain depot management and it is time-saving, not labor-intensive. This paper proposed a method of using a temperature contour map converted from digital temperature data to classify stored grain inventory modes in a large bulk grain warehouse, which mainly included detection of inventory changes and routine operations performed (aeration). The back propagation (BP) neural network was used in this method to identify and classify grain storage inventory modes based on the temperature contour map for helping grain depot management work. The method extracted and combined color coherence vector (CCV), texture feature vector (TFV) and smoothness feature vector (SFV) of temperature contour maps as the input vector of the BP neural network, and used inventory modes as the output vector. The experimental results indicated that the accuracy of the BP neural network with vector (CCV and TFV and SFV) as the input vector was about 93.9%, and its training time and prediction time were 320 and 0.12 s, respectively.

OWA aggregation with dual preferences for basic uncertain information

Basic Uncertain Information (BUI) as a newly introduced concept generalized a wide range of uncertain information. The well-known Ordered Weighted Averaging (OWA) operators can flexibly and effectively model bipolar preferences of decision makers over given real valued input vector. However, there are no extant methods for OWA operators to be carried out over given BUI vectors. Against this background, this study firstly discusses the interval transformation for BUI and elaborately explains the reasonability within it. Then, we propose the corresponding preference aggregations for BUI in two different decisional scenarios, the aggregation for BUI vector without original information influencing and the aggregation for BUI vector with original information influencing after interval transformation. For each decisional scenario, we also discuss two different orderings of preference aggregation, namely, interval-vector and vector-interval orderings, respectively. Hence, we will propose four different aggregation procedures of preference aggregation for BUI vector. Some illustrative examples are provided immediately after the corresponding aggregation procedures.

Modeling the trajectory of motion of a linear dynamic system with multi-point conditions

<abstract><p>The motion of the linear dynamic system with given properties is modeled; conditions for system state at various arbitrarily points in time are given. Simulated movement carried out due to the calculated input vector function. The method of undefined coefficients is used to construct the input vector function and the corresponding trajectory. The proposed method consists in the formation of the state vector function, the trajectory of motion and the input vector function in exponential-polynomial form, that is, in the form of linear combinations of the powers of the time parameter with vector coefficients. This linear combination is complemented by a scalar exponential function with an additional parameter in the exponent to change the type of trajectory. To find the introduced coefficients, formulas and a linear algebraic system are formed. To find the introduced coefficients, the formed linear combinations are substituted directly into the equations describing the dynamic system and into the given multipoint conditions for finding the entered coefficients. All this leads to obtaining algebraic formulas and linear algebraic systems. Only the matrices included in the system that describe the dynamics of the model (and similar matrices with higher exponents) are the coefficients for the unknown parameters of the resulting algebraic system. It is proved that the fulfillment of the condition Kalman is sufficient for the solvability of the resulting system. To substantiate the solvability of the system, the properties of finite-dimensional mappings are used: decomposition of spaces into subspaces, projectors on subspaces, semi-inverse operators. But for the practical use of the proposed method, it is sufficient to solve the obtained linear algebraic system and use the obtained linear formulas. The correctness of the obtained model is investigated. Due to the non-uniqueness of the solution to the problem posed, the trajectory of motion can be unstable. It is revealed which components of the desired coefficients are arbitrary. It is showed which ones to choose, to make the movement steady, that is, so that small changes in the given multi-point values, as well as a small change parameters of the dynamic system corresponded to a small change in the trajectory of motion. An example is given of constructing trajectories of a material point in a vertical plane under the action of a reactive force in order to hit a given point with a given speed.</p></abstract>

RUNOFF ESTIMATION IN URBAN CATCHMENT USING ARTIFICIAL NEURAL NETWORK MODELS

Many types of physical models have been developed for runoff estimation with successful results. However, accurate estimation of runoff remains a challenging problem owing to the lack of field data and the complexity of its hydrological process. In this paper, a machine learning method for runoff estimation is presented as an alternative approach to the physical model. Various types of input variables and artificial neural network (ANN) architectures were examined in this study. Results showed that a two-layer network with the tansig activation function and the Levenberg–Marquardt learning algorithm performed the best. For this architecture, the most effective input vector consists of a catchment perimeter, canal length, slope, runoff coefficient, and rainfall intensity. However, results of multivariate analysis of variance indicated the significant interaction effect of input data and the ANN architecture. Thus, to create a suitable ANN model for runoff estimation, a systematic determination of the input vector is necessary