Continuous Dimensions and Distributional Learning

2018 ◽  
pp. 127-154
Author(s):  
Vsevolod Kapatsinski
Keyword(s):  
Language Change ◽  
Irrelevant Dimension ◽  
Receptive Fields ◽  
Population Coding ◽  
Generative Model ◽  
Learning Mechanisms ◽  
Iterated Learning ◽  
Distributional Learning ◽  
The Difference ◽  
Attention Shifts

This chapter describes the evidence for the existence of dimensions, focusing on the difference between the difficulty of attention shifts to a previously relevant vs. irrelevant dimension. It discusses the representation of continuous dimensions in the associationist framework. including population coding and thermometer coding, as well as the idea that learning can adjust the breadth of adjustable receptive fields. In phonetics, continuous dimensions have been argued to be split into categories via distributional learning. This chapter reviews what we know about distributional learning and argues that it relies on several distinct learning mechanisms, including error-driven learning at two distinct levels and building a generative model of the speaker. The emergence of perceptual equivalence regions from error-driven learning is discussed, and implications for language change briefly noted with an iterated learning simulation.

scholarly journals A Stochastic Model for Block Segmentation of Images Based on the Quadtree and the Bayes Code for It

Entropy ◽  
2021 ◽  
Vol 23 (8) ◽  
pp. 991
Author(s):  
Yuta Nakahara ◽  
Toshiyasu Matsushima
Keyword(s):  
Computational Cost ◽  
Block Size ◽  
Input Image ◽  
Generative Model ◽  
Image Size ◽  
Variable Block ◽  
General Data ◽  
The Difference ◽  
Segmentation Of Images ◽  
Target Data

In information theory, lossless compression of general data is based on an explicit assumption of a stochastic generative model on target data. However, in lossless image compression, researchers have mainly focused on the coding procedure that outputs the coded sequence from the input image, and the assumption of the stochastic generative model is implicit. In these studies, there is a difficulty in discussing the difference between the expected code length and the entropy of the stochastic generative model. We solve this difficulty for a class of images, in which they have non-stationarity among segments. In this paper, we propose a novel stochastic generative model of images by redefining the implicit stochastic generative model in a previous coding procedure. Our model is based on the quadtree so that it effectively represents the variable block size segmentation of images. Then, we construct the Bayes code optimal for the proposed stochastic generative model. It requires the summation of all possible quadtrees weighted by their posterior. In general, its computational cost increases exponentially for the image size. However, we introduce an efficient algorithm to calculate it in the polynomial order of the image size without loss of optimality. As a result, the derived algorithm has a better average coding rate than that of JBIG.

Responses of Medullary Dorsal Horn Neurons to Corneal Stimulation by CO2 Pulses in the Rat

1999 ◽  
Vol 82 (5) ◽  
pp. 2092-2107 ◽  
Author(s):  
Harumitsu Hirata ◽  
James W. Hu ◽  
David A. Bereiter
Keyword(s):  
Thalamic Nucleus ◽  
Receptive Fields ◽  
Male Rats ◽  
Cervical Cord ◽  
Type I ◽  
Type Ii ◽  
Upper Cervical ◽  
The Difference ◽  
Evoked Activity ◽  
Stimulation Of

Corneal-responsive neurons were recorded extracellularly in two regions of the spinal trigeminal nucleus, subnucleus interpolaris/caudalis (Vi/Vc) and subnucleus caudalis/upper cervical cord (Vc/C1) transition regions, from methohexital-anesthetized male rats. Thirty-nine Vi/Vc and 26 Vc/C1 neurons that responded to mechanical and electrical stimulation of the cornea were examined for convergent cutaneous receptive fields, responses to natural stimulation of the corneal surface by CO2 pulses (0, 30, 60, 80, and 95%), effects of morphine, and projections to the contralateral thalamus. Forty-six percent of mechanically sensitive Vi/Vc neurons and 58% of Vc/C1 neurons were excited by CO2 stimulation. The evoked activity of most cells occurred at 60% CO2 after a delay of 7–22 s. At the Vi/Vc transition three response patterns were seen. Type I cells ( n = 11) displayed an increase in activity with increasing CO2 concentration. Type II cells ( n = 7) displayed a biphasic response, an initial inhibition followed by excitation in which the magnitude of the excitatory phase was dependent on CO2 concentration. A third category of Vi/Vc cells (type III, n = 3) responded to CO2 pulses only after morphine administration (>1.0 mg/kg). At the Vc/C1 transition, all CO2-responsive cells ( n = 15) displayed an increase in firing rates with greater CO2 concentration, similar to the pattern of type I Vi/Vc cells. Comparisons of the effects of CO2 pulses on Vi/Vc type I units, Vi/Vc type II units, and Vc/C1 corneal units revealed no significant differences in threshold intensity, stimulus encoding, or latency to sustained firing. Morphine (0.5–3.5 mg/kg iv) enhanced the CO2-evoked activity of 50% of Vi/Vc neurons tested, whereas all Vc/C1 cells were inhibited in a dose-dependent, naloxone-reversible manner. Stimulation of the contralateral posterior thalamic nucleus antidromically activated 37% of Vc/C1 corneal units; however, no effective sites were found within the ventral posteromedial thalamic nucleus or nucleus submedius. None of the Vi/Vc corneal units tested were antidromically activated from sites within these thalamic regions. Corneal-responsive neurons in the Vi/Vc and Vc/C1 regions likely serve different functions in ocular nociception, a conclusion reflected more by the difference in sensitivity to analgesic drugs and efferent projection targets than by the CO2 stimulus intensity encoding functions. Collectively, the properties of Vc/C1 corneal neurons were consistent with a role in the sensory-discriminative aspects of ocular pain due to chemical irritation. The unique and heterogeneous properties of Vi/Vc corneal neurons suggested involvement in more specialized ocular functions such as reflex control of tear formation or eye blinks or recruitment of antinociceptive control pathways.

scholarly journals Sin, hendes og KorpusDK

2017 ◽  
Vol 25 (24) ◽  
Author(s):  
Sten Vikner ◽  
Katrine Rosendal Ehlers
Keyword(s):  
Language Change ◽  
The Difference ◽  
The Moment ◽  
New Situation ◽  
National Corpus

We will briefly present our analysis of the four reflexive/non-reflexive types of pronouns in Danish (jf. Vikner 1985 and Ehlers & Vikner 2017), which each have to comply with two conditions: One type of condition on elements with/without selv (i.e. sig/hende vs. sig selv/ hende selv), and another type of condition on elements with/without sig (i.e. sig/sig selv vs. hende/hende selv). We will then show how the difference between the reflexive possessive and the non-reflexive possessives (sin vs. hendes) fits into this system, and we will examine how compatible our predictions are with data from the Danish national corpus, KorpusDK. We assume that sin corresponds to both sig and sig selv, and we argue against the view that there is a difference between sin and sin egen which corresponds to the difference between sig and sig selv. Finally we will discuss the language change that seems to be taking place at the moment, from a situation where sin may only have a singular antecedent to the new situation where sin may also have a plural antecedent.

Diurnal changes in retinula cell sensitivities and receptive fields (two-dimensional angular sensitivity functions) in the apposition eyes of Ligia exotica (Crustacea, Isopoda)

2001 ◽  
Vol 204 (2) ◽  
pp. 239-248 ◽  
Author(s):  
T. Hariyama ◽  
V.B. Meyer-Rochow ◽  
T. Kawauchi ◽  
Y. Takaku ◽  
Y. Tsukahara
Keyword(s):  
Receptive Fields ◽  
Resolving Power ◽  
Pigment Cells ◽  
Diurnal Changes ◽  
Two Dimensional ◽  
Angular Sensitivity ◽  
Pigment Granules ◽  
Screening Pigment ◽  
The Difference ◽  
Night And Day

The structural organization of the retinula cells in the eye of Ligia exotica changes diurnally. At night, the microvilli elongate, losing the regular and parallel alignment characteristic of the day condition. Crystalline cones and distal rhabdom tips are not pushed into each other during the day, but at night the rhabdoms protrude into the crystalline cones by up to 5 microm. Screening pigment granules in the retinula cells disperse during the night, but migrate radially towards the vicinity of the rhabdom during the day. No such displacements of the pigment granules of either distal or proximal screening pigment cells were observed. The sensitivity of the eye, monitored by electroretinogram (ERG) recordings, changes diurnally: values at midnight are, on average, 10 times those occurring during the day. However, intracellular recordings from single retinula cells (50 during the day and 50 at night) indicate that the difference between night and day sensitivities is only 2.5-fold. Two-dimensional angular sensitivity curves, indicative of a single unit's spatial sensitivity, had considerably less regular outlines at night than during the day. If based on the 50 % sensitivity level, day and night eyes possessed receptive fields of almost identical width (approximately 2 degrees), but if sensitivities below the 50 % limit were included, then receptive fields at night were significantly more extensive. We suggest that the morphological adaptations and diurnal changes in chromophore content seen in the apposition eye of L. exotica allow this animal to improve its photon capture at night while preserving at least some of the spatial resolving power characteristic of the light-adapted state. This would explain why this animal is capable of performing complex escape behaviours in the presence of predators both in bright and in very dim light.

Recovery of function in dorsal horn following partial deafferentation

1980 ◽  
Vol 43 (1) ◽  
pp. 102-117 ◽  
Author(s):  
L. M. Pubols ◽  
M. E. Goldberger
Keyword(s):  
Dorsal Horn ◽  
Recovery Of Function ◽  
Receptive Fields ◽  
Body Region ◽  
Partial Recovery ◽  
Collateral Sprouting ◽  
Somatotopic Organization ◽  
Tactile Input ◽  
Spinal Cord Segment ◽  
The Difference

1. Collateral sprouting of L6 dorsal root afferents within the dorsal horn of the L6 spinal cord segment has been shown anatomically to occur following transection of all other lumbosacral dorsal roots in the cat. The present study was performed to examine a possible physiological correlate of that sprouting, namely, an altered somatotopic organization of the dorsal horn at L6. This was evaluated by microelectrode mapping of the L6 dorsal horn in normal cats and in cats with L6 spared, lumbosacral dorsal rhizotomies performed 2 days (subacute spared root) or more than 8 wk (chronic spared root), prior to recording. 2. In normal cats the mediolateral somatotopic sequence of hindlimb representation in the L6 dorsal horn is ventral digits 2 and 3, dorsal digits 2 and 3, dorsal foot, rostral and lateral ankle, lateral leg, lateral thigh, and back. In both subacute and chronic spared-root cats the somatotopic sequence is similar to that of normal cats, but there is a loss of proximal thigh and back representation. This proximal body region is represented at the lateral edge of the dorsal horn in normal animals. 3. There was a partial loss of responsiveness of cells in the dorsal horn in the subacute spared-root group and a partial recovery of responsiveness in the chronic group. In the subacute group punctures exhibiting no responses to tactile input tended to be clustered in the lateral dorsal horn. 4. The lateral one-fourth of the dorsal horn in each animal was analyzed in terms of the percentage of recording loci occurring within it. The percentages of recording loci having receptive fields proximal to, distal to, and spanning the middle of the thigh (proximal, distal, and intermediate RFs) were tabulated for each animal. Subacute animals had a significantly lower-than-normal overall percentage of responsive loci in the lateral dorsal horn, but chronic animals did not. The percentage of distal fields therein was not different for the normal versus the subacute group, signifying that the loss of proximal and intermediate fields was responsible for the difference in overall percentage. Chronic animals, however, had significantly more distal fields than did normals. When all fields having any distal component were compared (i.e., distal and intermediate), the difference between the chronic and normal groups did not reach significance. One possible explanation of these findings is that loci having both proximal and distal RF components are unresponsive 2 days after partial denervation, but recover responsiveness to their spared distal input over an 8-wk period. One possible mechanism mediating these changes is localized sprouting of intact, spared axons. Other mechanisms of functional recovery, such as interneuronal sprouting, denervation supersensitivity, and unmasking of latent synapses, are discussed in relation to these and other data.

Visual response properties of neurons in four extrastriate visual areas of the owl monkey (Aotus trivirgatus): a quantitative comparison of medial, dorsomedial, dorsolateral, and middle temporal areas

1981 ◽  
Vol 45 (3) ◽  
pp. 397-416 ◽  
Author(s):  
J. F. Baker ◽  
S. E. Petersen ◽  
W. T. Newsome ◽  
J. M. Allman
Keyword(s):  
Receptive Fields ◽  
Field Size ◽  
Visual Response ◽  
Response Properties ◽  
Middle Temporal ◽  
Optimal Direction ◽  
Visual Areas ◽  
Direction Of Movement ◽  
Owl Monkeys ◽  
The Difference

1. The response properties of 354 single neurons in the medial (M), dorsomedial (DM), dorsolateral (DL), and middle temporal (MT) visual areas were studied quantitatively with bar, spot, and random-dot stimuli in chronically implanted owl monkeys with fixed gaze. 2. A directionality index was computed to compare the responses to stimuli in the optimal direction with the responses to the opposing direction of movement. The greater the difference between opposing directions, the higher the index. MT cells had much higher direction indices to moving bars than cells in DL, DM, and M. 3. A tuning index was computed for each cell to compare the responses to bars moving in the optimal direction, or flashed in the optimal orientation, with the responses in other directions or orientations within +/- 90 degrees. Cells in all four areas were more sharply tuned to the orientation of stationary flashed bars than to moving bars, although a few cells (9/92( were unresponsive in the absence of movement. DM cells tended to be more sharply tuned to moving bars than cells in the other areas. 4. Directionality in DM, DL, and MT was relatively unaffected by the use of single-spot stimuli instead of bars; tuning in all four areas was broader to spots than bars. 5. Moving arrays of randomly spaced spots were more strongly excitatory than bar stimuli for many neurons in MT (16/31 cells). These random-dot stimuli were also effective in M, but evoked no response or weak responses from most cells in DM and DL. 6. The best velocities of movement were usually in the range of 10-100 degrees/s, although a few cells (22/227), primarily in MT (14/69 cells), preferred higher velocities. 7. Receptive fields of neurons in all four areas were much larger than striate receptive fields. Eccentricity was positively correlated with receptive-field size (r = 0.62), but was not correlated with directionality index, tuning index, or best velocity. 8. The results support the hypothesis that there are specializations of function among the cortical visual areas.

scholarly journals Feature attention for binocular disparity in primate area MT depends on tuning strength

2015 ◽  
Vol 113 (5) ◽  
pp. 1545-1555 ◽  
Author(s):  
Douglas A. Ruff ◽  
Richard T. Born
Keyword(s):  
Binocular Disparity ◽  
Receptive Fields ◽  
Sensory Information ◽  
Neuronal Population ◽  
Area Mt ◽  
Temporal Area ◽  
Mt Neurons ◽  
Feature Attention ◽  
The Difference ◽  
The Relationship

Attending to a stimulus modulates the responses of sensory neurons that represent features of that stimulus, a phenomenon named “feature attention.” For example, attending to a stimulus containing upward motion enhances the responses of upward-preferring direction-selective neurons in the middle temporal area (MT) and suppresses the responses of downward-preferring neurons, even when the attended stimulus is outside of the spatial receptive fields of the recorded neurons (Treue S, Martinez-Trujillo JC. Nature 399: 575–579, 1999). This modulation renders the representation of sensory information across a neuronal population more selective for the features present in the attended stimulus (Martinez-Trujillo JC, Treue S. Curr Biol 14: 744–751, 2004). We hypothesized that if feature attention modulates neurons according to their tuning preferences, it should also be sensitive to their tuning strength, which is the magnitude of the difference in responses to preferred and null stimuli. We measured how the effects of feature attention on MT neurons in rhesus monkeys ( Macaca mulatta) depended on the relationship between features—in our case, direction of motion and binocular disparity—of the attended stimulus and a neuron's tuning for those features. We found that, as for direction, attention to stimuli containing binocular disparity cues modulated the responses of MT neurons and that the magnitude of the modulation depended on both a neuron's tuning preferences and its tuning strength. Our results suggest that modulation by feature attention may depend not just on which features a neuron represents but also on how well the neuron represents those features.

Are neurons in cat posteromedial lateral suprasylvian visual cortex orientation sensitive? Tests with bars and gratings

1995 ◽  
Vol 12 (1) ◽  
pp. 141-151 ◽  
Author(s):  
Yuri Danilov ◽  
Rodney J. Moore ◽  
Von R. King ◽  
Peter D. Spear
Keyword(s):  
Visual Cortex ◽  
Receptive Fields ◽  
Orientation Tuning ◽  
Orientation Sensitivity ◽  
Tuning Curves ◽  
Preferred Direction ◽  
Response Measures ◽  
Direction Of Movement ◽  
The Difference ◽  
Quantitative Response

AbstractThere is controversy in the literature concerning whether or not neurons in the cat's posteromedial lateral suprasylvian (PMLS) visual cortex are orientation selective. Previous studies that have tested cells with simple bar stimuli have found that few, if any, PMLS cells are orientation selective. Conversely, studies that have used repetitive stimuli such as gratings have found that most or all PMLS cells are orientation selective. It is not known whether this difference in results is due to the stimuli used or the laboratories using them. The present experiments were designed to answer this question by testing individual PMLS neurons for orientation sensitivity with both bar and grating stimuli. Using quantitative response measures, we found that most PMLS neurons respond well enough to stationary flashed stimuli to use such stimuli to test for orientation sensitivity. On the basis of these tests, we found that about 85% of the cells with well-defined receptive fields are orientation sensitive to flashed gratings, and a similar percentage are orientation sensitive to flashed bars. About 80% of the cells were orientation sensitive to both types of stimuli. The preferred orientations typically were similar for the two tests, and they were orthogonal to the preferred direction of movement. The strength of the orientation sensitivity (measured as the ratio of discharge to the preferred and nonpreferred orientations) was similar to both types of stimuli. However, the width of the orientation tuning curves was systematically broader to bars than to gratings. Several hypotheses are considered as to why previous studies using bars failed to find evidence for orientation sensitivity. In addition, a mechanism for the difference in orientation tuning to bars and gratings is suggested.

The Influence of Institutional Affiliation and Social Ecology on Sound Change

2020 ◽  
pp. 1-69
Author(s):  
Michael J. Fox
Keyword(s):  
Social Interaction ◽  
Social Ecology ◽  
Language Change ◽  
Speech Community ◽  
Social Mechanisms ◽  
Institutional Affiliation ◽  
Geographic Context ◽  
The Social ◽  
Geographic Regions ◽  
The Difference

The social mechanisms that influence the direction of language change operate along the demarcations of networks of communication (Bloomfield 1933; Milroy and Milroy 1985). Within geographic regions, the focused organizations that individuals participate in structure the lines of communication (Feld 1981) and the socio-demographic composition (social ecology) therein limits the options of peers to associate with (McPherson, Smith-Lovin, and Cook 2001). Schools have their own social ecology (McFarland et al. 2014) and attendance at schools can explain language change at a level above social interaction but below the level of community (Dodsworth and Benton 2017, 2019). This study uses acoustic vowel measurements from 132 speakers in three geographically contiguous cities located in northwestern Wisconsin. Modeling results indicate (1) similar socio-geographic contexts lead to linguistic similarity; (2) dissimilarity in social ecology leads to greater linguistic dissimilarity as the difference between a dyads’ years of birth increases; (3) net of local socio-geographic context and social ecology, similarity in sex and age leads to linguistic similarity and vice versa. These patterns indicate that local social ecologies further demarcate the lines of communication thereby structuring the form of language at a level between the micro interactional and the macro level of the speech community.

Multisensory Bayesian Inference Depends on Synapse Maturation during Training: Theoretical Analysis and Neural Modeling Implementation

2017 ◽  
Vol 29 (3) ◽  
pp. 735-782 ◽  
Author(s):  
Mauro Ursino ◽  
Cristiano Cuppini ◽  
Elisa Magosso
Keyword(s):  
Bayesian Inference ◽  
Maximum Likelihood ◽  
Maximum Likelihood Estimate ◽  
Experimental Studies ◽  
Receptive Fields ◽  
Learning Rule ◽  
Population Coding ◽  
Audiovisual Integration ◽  
Likelihood Estimate ◽  
Bayesian Estimate

Recent theoretical and experimental studies suggest that in multisensory conditions, the brain performs a near-optimal Bayesian estimate of external events, giving more weight to the more reliable stimuli. However, the neural mechanisms responsible for this behavior, and its progressive maturation in a multisensory environment, are still insufficiently understood. The aim of this letter is to analyze this problem with a neural network model of audiovisual integration, based on probabilistic population coding—the idea that a population of neurons can encode probability functions to perform Bayesian inference. The model consists of two chains of unisensory neurons (auditory and visual) topologically organized. They receive the corresponding input through a plastic receptive field and reciprocally exchange plastic cross-modal synapses, which encode the spatial co-occurrence of visual-auditory inputs. A third chain of multisensory neurons performs a simple sum of auditory and visual excitations. The work includes a theoretical part and a computer simulation study. We show how a simple rule for synapse learning (consisting of Hebbian reinforcement and a decay term) can be used during training to shrink the receptive fields and encode the unisensory likelihood functions. Hence, after training, each unisensory area realizes a maximum likelihood estimate of stimulus position (auditory or visual). In cross-modal conditions, the same learning rule can encode information on prior probability into the cross-modal synapses. Computer simulations confirm the theoretical results and show that the proposed network can realize a maximum likelihood estimate of auditory (or visual) positions in unimodal conditions and a Bayesian estimate, with moderate deviations from optimality, in cross-modal conditions. Furthermore, the model explains the ventriloquism illusion and, looking at the activity in the multimodal neurons, explains the automatic reweighting of auditory and visual inputs on a trial-by-trial basis, according to the reliability of the individual cues.

Sign in / Sign up

Export Citation Format

Share Document