人工神经网络与相容性生物量模型预测单木地上生物量的比较

森林生物量是林业生产经营和森林资源监测的重要指标,为探索高效低偏的单木生物量估测方法,引入人工神经网络.本研究采用黑龙江省东折棱河林场的101株长白落叶松地上生物量数据,基于不同变量(胸径、树高、冠幅)组合建立了4个聚合模型体系(AMS),采用加权回归消除模型的异方差.然后,基于最优的变量组合建立人工神经网络(ANN)...     

Optimization for high-density cultivation of heterotrophic Chlorella based on a hybrid neural network model

AIMS: The purpose of this study was to develop a reliable hybrid neural network (HNN) model for heterotrophic growth of Chlorella, based on which optimization for fed-batch (FB) cultivation of Chlorella may be successfully realized. METHODS AND RESULTS: Deterministic kinetic model was preliminarily developed for the optimization of FB cultivation of Chlorella. The highest biomass concentration and the maximum productivity were obtained as: 104.9 g l(-1) dry cell weight and 0.613 g l(-1) h(-1), respectively. After several cultivations had been performed, an HNN model was developed. The efficiency of biomass production was further increased by the optimization using this model. The highest biomass concentration and the maximum productivity attained was: 116.2 g l(-1) dry cell weight and 1.020 g l(-1) h(-1), respectively. CONCLUSION: The HNN model agreed well with experimental results in different cultivations. Comparison between the HNN model and the deterministic model showed that the former had better generalization ability, which made it a reliable tool in modelling and optimization. SIGNIFICANCE AND IMPACT OF THE STUDY: The high cell density and productivity of biomass obtained in this study is of significance for the commercial cultivation of Chlorella. The simple and efficient optimization strategy proposed in this paper may be employed in heterotrophic mass culture of Chlorella as well as other similar organisms.     

Computational antisense oligo prediction with a neural network model

MOTIVATION: The expression of a gene can be selectively inhibited by antisense oligonucleotides (AOs) targeting the mRNA. However, if the target site in the mRNA is picked randomly, typically 20% or less of the AOs are effective inhibitors in vivo. The sequence properties that make an AO effective are not well understood, thus many AOs need to be tested to find good inhibitors, which is time consuming and costly. So far computational models have been based exclusively on RNA structure prediction or motif searches while ignoring information from other aspects of AO design into the model. RESULTS: We present a computational model for AO prediction based on a neural network approach using a broad range of input parameters. Collecting sequence and efficacy data from AO scanning experiments in the literature generated a database of 490 AO molecules. Using a set of derived parameters based on AO sequence properties we trained a neural network model. The best model, an ensemble of 10 networks, gave an overall correlation coefficient of 0.30 (p=10(-8)). This model can predict effective AOs (>50% inhibition of gene expression) with a success rate of 92%. Using these thresholds the model predicts on average 12 effective AOs per 1000 base pairs, making it a stringent yet practical method for AO prediction.     

Bifurcation analysis of a neural network model

This paper describes the analysis of the well known neural network model by Wilson and Cowan. The neural network is modeled by a system of two ordinary differential equations that describe the evolution of average activities of excitatory and inhibitory populations of neurons. We analyze the dependence of the model's behavior on two parameters. The parameter plane is partitioned into regions of equivalent behavior bounded by bifurcation curves, and the representative phase diagram is constructed for each region. This allows us to describe qualitatively the behavior of the model in each region and to predict changes in the model dynamics as parameters are varied. In particular, we show that for some parameter values the system can exhibit long-period oscillations. A new type of dynamical behavior is also found when the system settles down either to a stationary state or to a limit cycle depending on the initial point.     

Control of bioreactors using a neural network model

Control of a continuous bioreactor based on a artificial neural network (ANN) model is carried out theoretically. The ANN model is identified, from input-output data of a bioreactor, using a three-layer feedforward network trained by a back propagation algorithm. The performance of the controller designed on the ANN model is compared with that of a conventional PI controller.     

Parameter identification of a model of yeast cultivation process with neural network

A mathematical model of yeast cultivation process has been proposed in [1]. The present paper describes procedure for parameter identification of this model. The obtained numerical values of the parameters of the model allow to estimate the specific growth rate μ.     

基于BP神经网络模型的平潭岛植被景观模拟

以GIS为技术平台,利用Matlab7.0,选用2km×2km网格图对平潭岛植被景观进行切割,得到50个样方数据,其中,44个用于模型训练,6个用于模型检验,在此基础上,选取分维数、Shannon多样性指数、蔓延度指数作为模型输出数据,选取居民点个数、风速和距海边距离作为影响因素,建立平潭岛植被景观的BP神经网络模型,并进行误差检验.结果表明:影响平潭岛植被景观空间格局和植被多样性状况的主要因素为风速、距海岸距离,人为因素对研究区植被景观的空间连接程度造成较大影响.BP神经网络模型对研究区植被景观与环境及人为影响因子之间关系的拟合与实际情况基本吻合,平均误差为7.4%,最小误差仅0.2%,模型模拟误差较小,拟合度较高,可用于对研究区植被景观的定量预测模拟.     

Predicting net joint moments during a weightlifting exercise with a neural network model

The purpose of this study was to develop and train a Neural Network (NN) that uses barbell mass and motions to predict hip, knee, and ankle Net Joint Moments (NJM) during a weightlifting exercise. Seven weightlifters performed two cleans at 85% of their competition maximum while ground reaction forces and 3-D motion data were recorded. An inverse dynamics procedure was used to calculate hip, knee, and ankle NJM. Vertical and horizontal barbell motion data were extracted and, along with barbell mass, used as inputs to a NN. The NN was then trained to model the association between the mass and kinematics of the barbell and the calculated NJM for six weightlifters, the data from the remaining weightlifter was then used to test the performance of the NN – this was repeated 7 times with a k-fold cross-validation procedure to assess the NN accuracy. Joint-specific predictions of NJM produced coefficients of determination (r²) that ranged from 0.79 to 0.95, and the percent difference between NN-predicted and inverse dynamics calculated peak NJM ranged between 5% and 16%. The NN was thus able to predict the spatiotemporal patterns and discrete peaks of the three NJM with reasonable accuracy, which suggests that it is feasible to predict lower extremity NJM from the mass and kinematics of the barbell. Future work is needed to determine whether combining a NN model with low cost technology (e.g., digital video and free digitising software) can also be used to predict NJM of weightlifters during field-testing situations, such as practice and competition, with comparable accuracy.     

A hybrid neural network of addressable and content-addressable memory

We investigate the memory structure and retrieval of the brain and propose a hybrid neural network of addressable and content-addressable memory which is a special database model and can memorize and retrieve any piece of information (a binary pattern) both addressably and content-addressably. The architecture of this hybrid neural network is hierarchical and takes the form of a tree of slabs which consist of binary neurons with the same array. Simplex memory neural networks are considered as the slabs of basic memory units, being distributed on the terminal vertexes of the tree. It is shown by theoretical analysis that the hybrid neural network is able to be constructed with Hebbian and competitive learning rules, and some other important characteristics of its learning and memory behavior are also consistent with those of the brain. Moreover, we demonstrate the hybrid neural network on a set of ten binary numeral patters     

On-line estimation of biomass concentration using a neural network and information about metabolic state

This paper deals with the design of a neural network-based biomass concentration estimation system. This system is enhanced by the incorporation of information about the actual metabolism of the microorganism cultivated, which is taken from an on-line knowledge-based system. Two different design approaches have been investigated using the fed-batch cultivation of bakers yeast as the model process. In the first, metabolic state (MS) data were passed as additional input to the neural network; in the second, these data were used to select a neural network suitable for the specific MS. Two neural network types—feed-forward (Levenberg-Marquardt) and cascade correlation—were applied to this system and tested, and the performances of these neural networks were compared.     

A discrete time neural network model with spiking neurons

We derive rigorous results describing the asymptotic dynamics of a discrete time model of spiking neurons introduced in Soula et al. (Neural Comput. 18, 1, 2006). Using symbolic dynamic techniques we show how the dynamics of membrane potential has a one to one correspondence with sequences of spikes patterns (“raster plots”). Moreover, though the dynamics is generically periodic, it has a weak form of initial conditions sensitivity due to the presence of a sharp threshold in the model definition. As a consequence, the model exhibits a dynamical regime indistinguishable from chaos in numerical experiments.      

Recognition of partly occluded patterns: a neural network model

 Human beings are often able to read a letter or word partly occluded by contaminating ink stains. However, if the stains are completely erased and the occluded areas of the letter are changed to white, we usually have difficulty in reading the letter. In this article I propose a hypothesis explaining why a pattern is easier to recognize when it is occluded by visible objects than by invisible opaque objects. A neural network model is constructed based on this hypothesis. The visual system extracts various visual features from the input pattern and then attempts to recognize it. If the occluding objects are not visible, the visual system will have difficulty in distinguishing which features are relevant to the original pattern and which are newly generated by the occlusion. If the occluding objects are visible, however, the visual system can easily discriminate between relevant and irrelevant features and recognize the occluded pattern correctly. The proposed model is an extended version of the neocognitron model. The activity of the feature-extracting cells whose receptive fields cover the occluding objects is suppressed in an early stage of the hierarchical network. Since the irrelevant features generated by the occlusion are thus eliminated, the model can recognize occluded patterns correctly, provided the occlusion is not so large as to prevent recognition even by human beings. Received: 21 February 2000 / Accepted in revised form: 11 September 2000     

Dynamics of a hybrid system of a brain neural network and an artificial nonlinear oscillator

In the brain, many functional modules interact with each other to execute complex information processing. Understanding the nature of these interactions is necessary for understanding how the brain functions. In this study, to mimic interacting modules in the brain, we constructed a hybrid system mutually coupling a hippocampal CA3 network as an actual brain module and a radial isochron clock (RIC) simulated by a personal computer as an artificial module. Return map analysis of the CA3-RIC system's dynamics showed the mutual entrainment and complex dynamics dependent on the coupling modes. The phase response curve of CA3 was modeled regarding the CA3 as a nonlinear oscillator. Using the phase response curves of CA3 and RIC, we reconstructed return maps of the hybrid system's dynamics. Although the reconstructed return maps almost agreed with the experimental data, there were deviations dependent on the coupling mode. In particular, we noted that the deviation was smaller under the bidirectional coupling conditions than during the one-way coupling from RIC to CA3. These results suggest that brain modules may flexibly change their dynamical properties through interaction with other modules.     

A neural network model of attention-modulated neurodynamics

Visual attention appears to modulate cortical neurodynamics and synchronization through various cholinergic mechanisms. In order to study these mechanisms, we have developed a neural network model of visual cortex area V4, based on psychophysical, anatomical and physiological data. With this model, we want to link selective visual information processing to neural circuits within V4, bottom-up sensory input pathways, top-down attention input pathways, and to cholinergic modulation from the prefrontal lobe. We investigate cellular and network mechanisms underlying some recent analytical results from visual attention experimental data. Our model can reproduce the experimental findings that attention to a stimulus causes increased gamma-frequency synchronization in the superficial layers. Computer simulations and STA power analysis also demonstrate different effects of the different cholinergic attention modulation action mechanisms.     

Coding proprioceptive information to control movement to a target: Simulation with a simple neural network

When the stick insect walks, the middle and rear legs step to positions immediately behind the tarsus of the adjacent rostral leg. Previous reports have described this movement to a target as a relationship between the tarsus positions of the two legs in a Cartesian coordinate system. However, leg proprioceptors measure the position of the target leg in terms of joint angles and leg muscles bring the tarsus of the moving leg to the proper end-point by establishing appropriate angles at the joints. Representation of this task in Cartesian coordinates requires non-linear coordinate transformations; realizing such a transformation in the nervous system appears to require many neurons. The present simulation using the back-propagation algorithm shows that a simple network of only nine units — 3 sensory input units, 3 motor output units, and 3 hidden units — suffices. The simulation also shows that an analytic coordinate transformation can be replaced by a direct association of joint configurations in the moving leg with those in the target leg.     

Transitory memory retrieval in a biologically plausible neural network model

A number of memory models have been proposed. These all have the basic structure that excitatory neurons are reciprocally connected by recurrent connections together with the connections with inhibitory neurons, which yields associative memory (i.e., pattern completion) and successive retrieval of memory. In most of the models, a simple mathematical model for a neuron in the form of a discrete map is adopted. It has not, however, been clarified whether behaviors like associative memory and successive retrieval of memory appear when a biologically plausible neuron model is used. In this paper, we propose a network model for associative memory and successive retrieval of memory based on Pinsky-Rinzel neurons. The state of pattern completion in associative memory can be observed with an appropriate balance of excitatory and inhibitory connection strengths. Increasing of the connection strength of inhibitory interneurons changes the state of memory retrieval from associative memory to successive retrieval of memory. We investigate this transition.     

Purposive behavior and cognitive mapping: a neural network model

This study presents a real-time, biologically plausible neural network approach to purposive behavior and cognitive mapping. The system is composed of (a) an action system, consisting of a goal-seeking neural mechanism controlled by a motivational system; and (b) a cognitive system, involving a neural cognitive map. The goal-seeking mechanism displays exploratory behavior until either (a) the goal is found or (b) an adequate prediction of the goal is generated. The cognitive map built by the network is a top logical map, i.e., it represents only the adjacency, but not distances or directions, between places. The network has recurrent and non-recurrent properties that allow the reading of the cognitive map without modifying it. Two types of predictions are introduced: fast-time and real-time predictions. Fast-time predictions are produced in advance of what occurs in real time, when the information stored in the cognitive map is used to predict the remote future. Real-time predictions are generated simultaneously with the occurrence of environmental events, when the information stored in the cognitive map is being updated. Computer simulations show that the network successfully describes latent learning and detour behavior in rats. In addition, simulations demonstrate that the network can be applied to problem-solving paradigms such as the Tower of Hanoi puzzle.