End-to-end face parsing via interlinked convolutional neural networks

Face parsing is an important computer vision task that requires accurate pixel segmentation of facial parts (such as eyes, nose, mouth, etc.), providing a basis for further face analysis, modification, and other applications. Interlinked Convolutional Neural Networks (iCNN) was proved to be an effective two-stage model for face parsing. However, the original iCNN was trained separately in two stages, limiting its performance. To solve this problem, we introduce a simple, end-to-end face parsing framework: STN-aided iCNN(STN-iCNN), which extends the iCNN by adding a Spatial Transformer Network (STN) between the two isolated stages. The STN-iCNN uses the STN to provide a trainable connection to the original two-stage iCNN pipeline, making end-to-end joint training possible. Moreover, as a by-product, STN also provides more precise cropped parts than the original cropper. Due to these two advantages, our approach significantly improves the accuracy of the original model. Our model achieved competitive performance on the Helen Dataset, the standard face parsing dataset. It also achieved superior performance on CelebAMask-HQ dataset, proving its good generalization. Our code has been released at https://github.com/aod321/STN-iCNN.     

Interfacing a flow cytometer computer with a PC-based patient information system

We have established an interface between our flow cytometer's computer and the personal computer (PC) which supports our patient database system. The PC has been equipped with a commercially available IEEE-488 bus interface board which is connected to the interface bus of the cytometer's Hewlett-Packard 9000/300 computer (HP). The PC is set as a bus device with the same address as that of the HP's printer. It is programmed to examine the stream of data sent to the printer and extract from it and store in an MS-DOS text file selected information which subsequently may be transferred to the database system.     

Synchronous and asynchronous brain-state-in-a-box information system neural models.

Synchronous and asynchronous information system neural models are proposed that are hybrids of Pawlak's information system and Brain-State-in-a-Box (BSB) neural models. The stability of the proposed models is studied using LaSalle's Invariance Principle. Applications to an analysis of the United Nations activities are presented as examples.     

Automatic identification of algae: neural network analysis of flow cytometric data

The performance of an artificial neural network for automaticidentification of phytoplankton was investigated with data fromalgal laboratory cultures, analysed on the Optical PlanktonAnalyser (OPA), a flow cytometer especially developed for theanalysis of phytoplankton. Data from monocultures of eight algalspecies were used to train a neural network. The performanceof the trained network was tested with OPA data from mixturesof laboratory cultures. The network could distinguish Cyanobacteriafrom other algae with 99% accuracy. The identification of specieswas performed with less accuracy, but was generally >90%.This indicates that a neural network under supervised learningcan be used for automatic identification of species in relativelycomplex mixtures. Incorporation of such a system may also increasethe operational size range of a flow cytometer. The combinationof the OPA and neural network data analysis offers the elementsto build an operational automatic algal identification system.     

Nonlinear system identification and control based on modular neural networks

A new approach for nonlinear system identification and control based on modular neural networks (MNN) is proposed in this paper. The computational complexity of neural identification can be greatly reduced if the whole system is decomposed into several subsystems. This is obtained using a partitioning algorithm. Each local nonlinear model is associated with a nonlinear controller. These are also implemented by neural networks. The switching between the neural controllers is done by a dynamical switcher, also implemented by neural networks, that tracks the different operating points. The proposed multiple modelling and control strategy has been successfully tested on simulated laboratory scale liquid-level system.     

Partial information decomposition reveals that synergistic neural integration is greater downstream of recurrent information flow in organotypic cortical cultures

The directionality of network information flow dictates how networks process information. A central component of information processing in both biological and artificial neural networks is their ability to perform synergistic integration–a type of computation. We established previously that synergistic integration varies directly with the strength of feedforward information flow. However, the relationships between both recurrent and feedback information flow and synergistic integration remain unknown. To address this, we analyzed the spiking activity of hundreds of neurons in organotypic cultures of mouse cortex. We asked how empirically observed synergistic integration–determined from partial information decomposition–varied with local functional network structure that was categorized into motifs with varying recurrent and feedback information flow. We found that synergistic integration was elevated in motifs with greater recurrent information flow beyond that expected from the local feedforward information flow. Feedback information flow was interrelated with feedforward information flow and was associated with decreased synergistic integration. Our results indicate that synergistic integration is distinctly influenced by the directionality of local information flow.     

A neural network study on the dynamic identification of a fermentation system

The objective of this paper is to propose neural networks for the study of dynamic identification and prediction of a fermentation system which produces mainly 2,3-butanediol (2,3-BDL). The metabolic products of the fermentation, acetic acid, acetoin, ethanol, and 2,3-BDL were measured on-line via a mass spectrometer modified by the insertion of a dimethylvinylsilicone membrane probe. The measured data at different sampling times were included as the input and output nodes, at different learning batches, of the network. A fermentation system is usually nonlinear and dynamic in nature. Measured fermentation data obtained from the complex metabolic pathways are often difficult to be entirely included in a static process model, therefore, a dynamic model was suggested instead. In this work, neural networks were provided by a dynamic learning and prediction process that moved along the time sequence batchwise. In other words, a scheme of two-dimensional moving window (number of input nodes by the number of training data) was proposed for reading in new data while forgetting part of the old data. Proper size of the network including proper number of input/output nodes were determined by trained with the real-time fermentation data. Different number of hidden nodes under the consideration of both learning performance and computation efficiency were tested. The data size for each learning batch was determined. The performance of the learning factors such as the learning coefficient η and the momentum term coefficient α were also discussed. The effect of different dynamic learning intervals, with different starting points and the same ending point, both on the learning and prediction performance were studied. On the other hand, the effect of different dynamic learning intervals, with the same starting point and different ending points, was also investigated. The size of data sampling interval was also discussed. The performance from four different types of transfer functions, x/(1+|x|), sgn(x)·x ²/(1+x ²), 2/(1+e ^{? x} )?1, and 1/(1+e ^{? x} ) was compared. A scaling factor b was added to the transfer function and the effect of this factor on the learning was also evaluated. The prediction results from the time-delayed neural networks were also studied.     

Positional information in neural map development: lessons from the olfactory system

Positional information is fundamental in development. Although molecular gradients are thought to represent positional information in various systems, the molecular logic used to interpret these gradients remains controversial. In the nervous system, sensory maps are formed in the brain based on gradients of axon guidance molecules. However, it remains unclear how axons find their targets based on relative, not absolute, expression levels of axon guidance receptors. No model solely based on axon-target interactions explains this point. Recent studies in the olfactory system suggested that the neural map formation requires axon-axon interactions, which is known as axon sorting. This review discusses how axon-axon and axon-target interactions interpret molecular gradients and determine the axonal projection sites in neural map formation.     

A polar co-ordinate system for positional information in the vertebrate neural retina

Pattern formation in many developing systems has been traditionally understood in terms of a prior signalling of positional information along mutually orthogonal axes, thus setting up a Cartesian co-ordinate grid. Existing data from the vertebrate neural retina has here been reinterpreted in terms of a non-Cartesian system. Results bearing on regeneration and duplication in eye fragments, “axial determination” and polarity alterations in fused eye fragments are considered and interpreted in terms of a polar co-ordinate system. On this model the position of each region of the retina is defined by a co-ordinate (radial) expressing distance from the centre and another (circumferential) expressing position around the circumference. The radial co-ordinate accords with the radial nature of retinal growth and it is possible that the spatially ordered sequence of cell divisions may itself specify the sequence of radial positions. The circumferential co-ordinate, unlike the radial counterpart, is specified in interaction with the extra-ocular tissue such that it may be oriented in the embryo with reference to a primary embryo Cartesian grid. The ability of the model to account for ultrastructural findings, problematic for the Cartesian model, is discussed.     

Information loss with transform methods in system identification: a new set of transforms with high information content

Linear transform methods like moments, modulating functions, and Laplace transforms are widely used for parameter estimation in system identification problems because they can reduce a large set of overdetermined equations to a small set of linear and nonlinear equations, which often have a very simple form and a unique solution. However, the effects of noise in the data are neglected in deriving these equations. We show (in terms of Fisher's information measure, the generalized variance, and simulations) that these methods can lead to very large errors in the estimates. We develop a new set of transforms based on the idea of maximizing their Fisher information content. The robustness of these new transforms, in contrast to the others, is illustrated by simulations of nanosecond flourescence decay and multicomponent exponential decay.     

Decreased thalamo-cortical connectivity by alteration of neural information flow in theta oscillation in depression-model rats

Alterations in oscillatory brain activity are strongly correlated with cognitive performance in various physiological rhythms. The present study investigated whether the directionality of neural information flow (NIF) could be used to characterize the synaptic plasticity in thalamocortical (TC) pathway, and examined which frequency field oscillations were mostly related to the cognitive deficiency in depression. Two novel algorithms were employed to determine the coupling interaction between the LD thalamus and medial prefrontal cortex (mPFC) in five frequency bands, using the phase signals of local field potentials (LFP) in these two regions. The results showed that the power of neural activity in mPFC was increased in delta, theta and beta frequency bands in depression. However, the nonlinear characteristics of LFP activity were weakened in depression by means of sample entropy measurements. In the analysis of phase dynamics, the phase synchronization values were reduced in theta rhythm in stressed rats. Importantly, the coupling direction index d and the unidirectional influence from LD thalamus to mPFC were significantly reduced at the theta rhythm in rats in depression, and increased after memantine treatment, which were associated with the LTP alterations and cognitive impairment in our previous report. Moreover, the fact that the reduced entropy value was only found in mPFC might implicate postsynaptic effect involved in synaptic plasticity alteration in the depression model. The results suggest that the effects of depression on cognitive deficits are mediated via profound alterations in information flow in the TC pathway, and the directional index at theta rhythm could be used as a measurement of synaptic plasticity.     

人工神经原网络处理PCR数据在细菌鉴定中的应用

目的16SrRNA和16S-23SrRNA间区片段是常用细菌分类鉴定靶点,本研究探讨人工神经原网络（ANN）对上述位点PCR扩增产物数据分析在细菌快速鉴定方面的价值。方法2对15SrRNA基因荧光引物和1对16S-23SrRNA区间基因引物用于扩增血液标本中分离出的317株细菌。相关毛细管电泳（CE）限制性片段长度多态性（RFLP）和单链构象多态性（SSCP）数据进行人工神经原网络分析。结果16S-23SrRNA基因的RFLP数据对未知菌鉴定的准确率高于16SrRNA基因的SSCP数据,分别为98．0％和79．6％。结论实验证明了人工神经原网络作为一种模式识别方法对于简化细菌鉴定十分有价值。     

Pollen texture identification using neural networks

The importance of research leading to the automation of pollen identification is briefly outlined. A new technique, neural network analysis, is briefly introduced, and then applied to the determination of light microscope images of pollen grains. The results are compared with some previously published statistical classifiers. Although both types of classifiers may work, the neural network is apparently superior to the statistical methods in three ways: high success rates (100% in this case), small number of samples needed for training, and simplicity of features.     

On the identification of neural responses

The most common form of measuring electrical responses of nerve cells is the recording of a given cell's spike train profile to the parameters of a given input signal. In this paper we consider the conditions under which it is possible to relate such response measures to (a) the properties of the cell's underlying activity characteristics, (b) the neural network, and (c) the input signal.This work was partially supported by the Natural Sciences and Engineering Research Council of Canada under Grant A-4345 to M.N.O. and under Grant A-4395 to T.M.C. through the University of Alberta     

与突触可塑性相关的神经振荡和信息流研究

作为一种有节律的神经活动,神经振荡现象发生在所有的神经系统中,例如大脑皮层、海马、皮层下神经核团以及感觉器官.本综述首先给出了已有的研究结果,即基于theta和gamma频段的同步神经振荡揭示了认知过程的起源与本质,如学习与记忆.然后介绍了关于神经振荡分析的新技术和算法,如表征神经元突触可塑性的神经信息流方向指数,并例...     

Aperiodic stochastic resonance in neural information processing with Gaussian colored noise

The aim of this paper is to explore the phenomenon of aperiodic stochastic resonance in neural systems with colored noise. For nonlinear dynamical systems driven by Gaussian colored noise, we prove that the stochastic sample trajectory can converge to the corresponding deterministic trajectory as noise intensity tends to zero in mean square, under global and local Lipschitz conditions, respectively. Then, following forbidden interval theorem we predict the phenomenon of aperiodic stochastic resonance in bistable and excitable neural systems. Two neuron models are further used to verify the theoretical prediction. Moreover, we disclose the phenomenon of aperiodic stochastic resonance induced by correlation time and this finding suggests that adjusting noise correlation might be a biologically more plausible mechanism in neural signal processing.