Robert Proulx

NDRAM: nonlinear dynamic recurrent associative memory for learning bipolar and nonbipolar correlated patterns

This paper presents a new unsupervised attractor neural network, which, contrary to optimal linear associative memory models, is able to develop nonbipolar attractors as well as bipolar attractors. Moreover, the model is able to develop less spurious attractors and has a better recall performance under random noise than any other Hopfield type neural network. Those performances are obtained by a simple Hebbian/anti-Hebbian online learning rule that directly incorporates feedback from a specific nonlinear transmission rule. Several computer simulations show the models distinguishing properties.

Behavioral avoidance dynamics in the presence of a virtual spider

Tracking behavior with a virtual spider and a neutral target is compared in fearful and nonfearful subjects. Head-tracking in virtual environments appears to be a scale-free behavior with long-range fractal-like patterns. Moreover, these fractal patterns change according to what the target affords the tracker and the level of behavioral avoidance manifested by the subjects. Results are interpreted in terms of ecological psychology and nonlinear dynamics, and implications for virtual reality (VR) psychology are outlined.

Particle swarm classification: A survey and positioning

This paper offers a survey of recent work on particle swarm classification (PSC), a promising offshoot of particle swarm optimization (PSO), with the goal of positioning it in the overall classification domain. The richness of the related literature shows that this new classification approach may be an efficient alternative, in addition to existing paradigms. After describing the various PSC approaches found in the literature, the paper identifies and discusses two data-related problems that may affect PSC efficiency: high-dimensional datasets and mixed-attribute data. The solutions that have been proposed in the literature for each of these issues are described including recent improvements by a novel PSC algorithm developed by the authors. Subsequently, a positioning PSC for these problems with respect to other classification approaches is made. This is accomplished by using one proprietary and five well known benchmark datasets to determine the performances of PSC algorithm and comparing the obtained results with those reported for various other classification approaches. It is concluded that PSC can be efficiently applied to classification problems with large numbers of instances, both in continuous and mixed-attribute problem description spaces. Moreover, the obtained results show that PSC may not only be applied to more demanding problem domains, but it can also be a competitive alternative to well established classification techniques.

Categorization in unsupervised neural networks: the Eidos model

Proulx and Begin (1995) recently explained the power of a learning rule that combines Hebbian and anti-Hebbian learning in unsupervised auto-associative neural networks. Combined with the brain-state-in-a-box transmission rule, this learning rule defines a new model of categorization: the Eidos model. To test this model, a simulated neural network, composed of 35 interconnected units, is subjected to an alphabetical characters recognition task. The results indicate the necessity of adding two parameters to the model: a restraining parameter and a forgetting parameter. The study shows the outstanding capacity of the model to categorize highly altered stimuli after a suitable learning process. Thus, the Eidos model seems to be an interesting option to achieve categorization in unsupervised neural networks.

Processing of pragmatic and facial affective information by patients with closed-head injuries

Although several affective impairments have been demonstrated to occur following closed-head injury (CHI), deficits of the communicative function of language, particularly sentenial and suprasentential pragmatic aspects, have been suggested, but not demonstrated, to occur. This study compared 31 normals and 31 severely closed-head injured patients matched for age, sex and education. The dependent measures consisted of a facial test of emotion (FTE) and a contextual test of emotion (CTE). The former task consisted of 36 slides representing facial expressions of the six emotions demonstrated by Ekman and colleagues to be transcultural, namely, job, sadness, fear, anger, surprise and disgust. The subject was required to name the appropriate emotion for each slide. The latter task consisted of correctly identifying the appropriate emotion for each of 36 brief verbal narratives representing contexts connotative of the same six transcultural emotions. The CHI patients were impaired overall on the FTE but not the CTE. However, the ability to identify anger was significantly impaired on both tasks when considered in isolation from the other emotions. It was concluded that a processing deficit of primary emotional material, particularly anger, does exist following CHI, but that this deficit is not necessarily independent of task and/or modality parameters. It was also concluded that evidence of a pragmatic deficit of the language function following CHI remains to be provided at this time.

FEBAM: A Feature-Extracting Bidirectional Associative Memory

In this paper, a new model that can ultimately create its own set of perceptual features is proposed. Using a bidirectional associative memory (BAM)-inspired architecture, the resulting model inherits properties such as attractor-like behavior and successful processing of noisy inputs, while being able to achieve principal component analysis (PCA) tasks such as feature extraction and dimensionality reduction. The model is tested by simulating image reconstruction and blind source separation tasks. Simulations show that the model fares particularly well compared to current neural PCA and independent component analysis (ICA) algorithms. It is argued the model possesses more cognitive explanative power than any other nonlinear/linear PCA and ICA algorithm.

Bottom-up learning of explicit knowledge using a Bayesian algorithm and a new Hebbian learning rule

The goal of this article is to propose a new cognitive model that focuses on bottom-up learning of explicit knowledge (i.e., the transformation of implicit knowledge into explicit knowledge). This phenomenon has recently received much attention in empirical research that was not accompanied by a corresponding work effort in cognitive modeling. The new model is called TEnsor LEarning of CAusal STructure (TELECAST). In TELECAST, implicit processing is modeled using an unsupervised connectionist network (the Joint Probability EXtractor: JPEX) while explicit (causal) knowledge is implemented using a Bayesian belief network (which is built online using JPEX). Every task is simultaneously processed explicitly and implicitly and the results are integrated to provide the model output. Here, TELECAST is used to simulate a causal inference task and two serial reaction time experiments.

Emotional Facial Expressive and Discriminative Performance and Lateralization in Normal Young Adults

Studies of cerebral dominance for posed emotional facial expression using free-viewing of hemicomposites have produced inconclusive findings, and the concordance of facial emotion identification (discrimination) and the expression of the same facial emotion remains unknown. Expressive and discriminative (14 men, 14 women) facial emotion performances of undergraduates and the lateralization of full-face and lower-face hemicomposite photographic montages of the expressions of six transcultural emotions (joy, sadness, fear, surprise, disgust, anger) as ascertained by 15 male and 15 female undergraduate judges were analyzed. All groups were matched for age and education. The lower face was non-significantly left-face dominant, sadness was strongly significantly right-face dominant and fear was non-significantly left face dominant. Both sexes were equally lateralized overall and demonstrated the same pattern as described above, though slight (apparently trivial) differences appeared in multivariate analysis, and in univariate interactions. Results were interpreted as non-supportive of a simple right hemisphere dominance model of facial affect, nor of a left-hemisphere-negative/right-hemisphere-positive model. It was concluded that facial affect dominance results are coherent only within, and not between, methods such as free viewing hemicomposite and tachistoscopic methods, and tasks, such as expressive and discriminative tasks.

Are unsupervised neural networks ignorant? Sizing the effect of environmental distributions on unsupervised learning

Learning environmental biases is a rational behavior: by using prior odds, Bayesian networks rapidly became a benchmark in machine learning. Moreover, a growing body of evidence now suggests that humans are using base rate information. Unsupervised connectionist networks are used in computer science for machine learning and in psychology to model human cognition, but it is unclear whether they are sensitive to prior odds. In this paper, we show that hard competitive learners are unable to use environmental biases while recurrent associative memories use frequency of exemplars and categories independently. Hence, it is concluded that recurrent associative memories are more useful than hard competitive networks to model human cognition and have a higher potential in machine learning.

A new online unsupervised learning rule for the BSB model

In this paper it is demonstrated that a new unsupervised learning rule enable a nonlinear model, like the BSB model and the Hopfield network, to learn online correlated stimuli. This rule stabilizes the weight matrix growth to the projection rule in a local fashion. The model has been tested with computer simulations that show that the model is stable over the variations of its free parameters and that it is noise tolerant in the recall task.