A neural network model of implicit memory for object recognition

People name well-known objects shown in pictures more quickly if they have studied them previously. The most common interpretation of this priming effect is that processing is facilitated by an implicit memory trace in a perceptual representation system. We show that object priming can be explained instead as a bias in information processing, without recourse to an implicit memory system. Assumptions about psychological decision-making processes and bias were added to a neural network model for object identification, and the model accounted for performance both qualitatively and quantitatively in four object identification experiments.     

Efficiency investigation from shallow to deep neural network techniques in human activity recognition

In the last years, several researchers measured different recognition rates with different artificial neural network (ANN) techniques on public data sets in the human activity recognition (HAR) problem. However an overall investigation does not exist in the literature and the efficiency of complex and deeper ANNs over shallow networks is not clear. The purpose of this paper is to investigate the recognition rate and time requirement of different kinds of ANN approaches in HAR. This work examines the performance of shallow ANN architectures with different hyper-parameters, ANN ensembles, binary ANN classifier groups, and convolutional neural networks on two public databases. Although the popularity of binary classifiers, classifier ensembles and deep learning have been significantly increasing, this study shows that shallow ANNs with appropriate hyper-parameters in combination with extracted features can reach similar or higher recognition rate in less time than other artificial neural network methods in HAR. With a well-tuned ANN we outperformed all previous results on two public databases. Consequently, instead of the more complex ANN techniques, the usage of simple ANN with two or three layers can be an appropriate choice for activity recognition.     

A neural basis for expert object recognition

Although most adults are considered to be experts in the identification of faces, fewer people specialize in the recognition of other objects, such as birds and dogs. In this research, the neurophysiological processes associated with expert bird and dog recognition were investigated using event-related potentials. An enhanced early negative component (N170, 164 ms) was found when bird and dog experts categorized objects in their domain of expertise relative to when they categorized objects outside their domain of expertise. This finding indicates that objects from well-learned categories are neurologically differentiated from objects from lesser-known categories at a relatively early stage of visual processing.     

Large-scale neural network for sentence processing

Our model of sentence comprehension includes at least grammatical processes important for structure-building, and executive resources such as working memory that support these grammatical processes. We hypothesized that a core network of brain regions supports grammatical processes, and that additional brain regions are activated depending on the working memory demands associated with processing a particular grammatical feature. We used functional magnetic resonance imaging (fMRI) to test this hypothesis by comparing cortical activation patterns during coherence judgments of sentences with three different syntactic features. We found activation of the ventral portion of left inferior frontal cortex during judgments of violations of each grammatical feature. Increased recruitment of the dorsal portion of left inferior frontal cortex was seen during judgments of violations of specific grammatical features that appear to involve a more prominent working memory component. Left posterolateral temporal cortex and anterior cingulate were also implicated in judging some of the grammatical features. Our observations are consistent with a large-scale neural network for sentence processing that includes a core set of regions for detecting and repairing several different kinds of grammatical features, and additional regions that appear to participate depending on the working memory demands associated with processing a particular grammatical feature.     

A neural network simulator for supercomputers

Connectionist modeling is computationally intensive. Until parallel computers become more widely available, supercomputing resources can be exploited. This paper describes a neural network simulator (NNS) written in FORTRAN for supercomputers. The present simulation engine consists of code for the backpropagation method of changing weights in connectionist models. A file interface reports simulation results in a variety of formats. The file interface also contains an interpreter for an input file through which the network structure is defined, the problem is represented, and various parameters are set. The input-file syntax is described in detail. NNS has been used both as an instructional aid and as a research tool. A simulation of “recovery of unrehearsed associations” is used to illustrate the use of the input file and to demonstrate the performance of NNS. Versions of NNS have been written for the Cray X-MP/48 and for the IBM 3090.     

Influence of movement activities on shape recognition and retention.

Three groups of 4-yr.-old children were asked to complete a form perception assessment instrument prior to, 1 hr. after, and 1 wk. following a treatment. Group 1 participated in a movement-based form perception program, while Group 2 was instructed using a traditional classroom method. A third group which acted as control participated in unrelated movement activities. A repeated-measures analysis of variance gave a main effect of tests and an interaction of groups X tests. Group 2 displayed significantly higher performance on the posttest than Group 1; however, after 7 wk. the performance of Group 2 had decreased to a level below that of Groups 1 and 3, which remained stable.     

Short-term memory for serial order: a recurrent neural network model

Despite a century of research, the mechanisms underlying short-term or working memory for serial order remain uncertain. Recent theoretical models have converged on a particular account, based on transient associations between independent item and context representations. In the present article, the authors present an alternative model, according to which sequence information is encoded through sustained patterns of activation within a recurrent neural network architecture. As demonstrated through a series of computer simulations, the model provides a parsimonious account for numerous benchmark characteristics of immediate serial recall, including data that have been considered to preclude the application of recurrent neural networks in this domain. Unlike most competing accounts, the model deals naturally with findings concerning the role of background knowledge in serial recall and makes contact with relevant neuroscientific data. Furthermore, the model gives rise to numerous testable predictions that differentiate it from competing theories. Taken together, the results presented indicate that recurrent neural networks may offer a useful framework for understanding short-term memory for serial order.     

Looking for a structural network: effects of changing size and style on letter recognition.

The accuracy of identifying backward-masked strings of unrelated letters was measured to test a prediction of a structural network model in which information about the absolute size of letters in a font is factored out from other types of information and is represented by a single size parameter. In this model, variation from block to block in the font size should be handled relatively easily by changing this parameter. In contrast, variation in several aspects of the font style would require changing multiple parameters and should be more difficult. In three experiments, the font was alternated in style, size, size and style, or not at all, at the end of each block of eight trials. The accuracy was lower overall when the font was varied, but contrary to the predictions of the model, variation in size reduced the accuracy of performance as much as did variation in style and variation in size and style.     

Model for a flexible motor memory based on a self-active recurrent neural network

Using recent recurrent network architecture based on the reservoir computing approach, we propose and numerically simulate a model that is focused on the aspects of a flexible motor memory for the storage of elementary movement patterns into the synaptic weights of a neural network, so that the patterns can be retrieved at any time by simple static commands. The resulting motor memory is flexible in that it is capable to continuously modulate the stored patterns. The modulation consists in an approximately linear inter- and extrapolation, generating a large space of possible movements that have not been learned before. A recurrent network of thousand neurons is trained in a manner that corresponds to a realistic exercising scenario, with experimentally measured muscular activations and with kinetic data representing proprioceptive feedback. The network is “self-active” in that it maintains recurrent flow of activation even in the absence of input, a feature that resembles the “resting-state activity” found in the human and animal brain. The model involves the concept of “neural outsourcing” which amounts to the permanent shifting of computational load from higher to lower-level neural structures, which might help to explain why humans are able to execute learned skills in a fluent and flexible manner without the need for attention to the details of the movement.     

How environmental movement constraints shape the neural code for space

Study of the neural code for space in rodents has many insights to offer for how mammals, including humans, construct a mental representation of space. This code is centered on the hippocampal place cells, which are active in particular places in the environment. Place cells are informed by numerous other spatial cell types including grid cells, which provide a signal for distance and direction and are thought to help anchor the place cell signal. These neurons combine self-motion and environmental information to create and update their map-like representation. Study of their activity patterns in complex environments of varying structure has revealed that this "cognitive map" of space is not a fixed and rigid entity that permeates space, but rather is variably affected by the movement constraints of the environment. These findings are pointing toward a more flexible spatial code in which the map is adapted to the movement possibilities of the space. An as-yet-unanswered question is whether these different forms of representation have functional consequences, as suggested by an enactivist view of spatial cognition.

     

Three-channel convolutional neural networks for vegetable leaf disease recognition

The color information of diseased leaf is the main basis for leaf based plant disease recognition. To make use of color information, a novel three-channel convolutional neural networks (TCCNN) model is constructed by combining three color components for vegetable leaf disease recognition. In the model, each channel of TCCNN is fed by one of three color components of RGB diseased leaf image, the convolutional feature in each CNN is learned and transmitted to the next convolutional layer and pooling layer in turn, then the features are fused through a fully connected fusion layer to get a deep-level disease recognition feature vector. Finally, a softmax layer makes use of the feature vector to classify the input images into the predefined classes. The proposed method can automatically learn the representative features from the complex diseased leaf images, and effectively recognize vegetable diseases. The experimental results validate that the proposed method outperforms the state-of-the-art methods of the vegetable leaf disease recognition.     

Outline shape is a mediator of object recognition that is particularly important for living things

We assess the importance of outline shape in mediating the recognition of living and nonliving things. Natural objects were presented as shaded line drawings or silhouettes, and were living and nonliving things. For object decision (deciding whether an object may be encountered in real life) there were longer response times to nonliving than to living things. Importantly, this category difference was greater for silhouettes than for shaded line drawings. For naming, similar category and stimulus differences were evident, but were not as pronounced. We also examined effects of prior naming on subsequent object decision performance. Repetition priming was equivalent for nonliving and living things. However, prior presentation of silhouettes (but not shaded line drawings) reduced the longer RT to nonliving things relative to living things in silhouette object decision. We propose that outline contour benefits recognition of living things more than nonliving things: For nonliving things, there may be greater 2-D/3-D interpretational ambiguity, and/or they may possess fewer salient features.     

A neural network approach to identifying adolescent adjustment.

This study examined the relationship between the quality of adjustment in adolescents and a set of psychiatric diagnoses, personality traits, parental bonding, and social support variables. One hundred fifty adolescents were administered the Millon Adolescent Personality Inventory, the Parental Bonding Questionnaire, the Social Support Questionnaire, and the Diagnostic Interview for Children and Adolescents. A neural network approach was then utilized, and it was found that several of the variables (e.g., Major Depressive Disorder, Conduct Disorder, and Societal Conformity) had a significant role in classifying adolescents into three groups: maladjusted, nominally adjusted, and well-adjusted.     

Posterior parietal cortex as part of a neural network for directed attention in rats

A rodent model of directed attention has been developed based upon behavioral analysis of contralateral neglect, pharmacological manipulations, and anatomical analysis of neural circuitry. In each of these three domains the rodent model exhibits striking similarities to humans. We hypothesize that there is a specific thalamo-cortical-basal ganglia network that subserves spatial attentional functions. Key components of this network are medial agranular and posterior parietal cortex, dorsocentral striatum, and the lateral posterior thalamic nucleus. Several issues need to be addressed before we can hope to realistically understand or model the functions of this network. Among these are the roles of medial versus lateral posterior parietal cortex; cholinergic mechanisms in attention; interhemispheric interactions; the role of synchronous firing at the cortical, striatal, and thalamic levels; interactions between cortical and thalamic projections to the striatum; interactions between cortical and nigral inputs to the thalamus; the role of collicular inputs to the lateral posterior thalamic nucleus; the role of cerebral cortex versus superior colliculus in driving the motor output expressed as orienting behavior during directed attention; the extent to which the circuitry we describe for directed attention also plays a role in other forms of attention.