nips nips2006 nips2006-145 knowledge-graph by maker-knowledge-mining
Source: pdf
Author: Jason M. Samonds, Brian R. Potetz, Tai S. Lee
Abstract: Although there has been substantial progress in understanding the neurophysiological mechanisms of stereopsis, how neurons interact in a network during stereo computation remains unclear. Computational models on stereopsis suggest local competition and long-range cooperation are important for resolving ambiguity during stereo matching. To test these predictions, we simultaneously recorded from multiple neurons in V1 of awake, behaving macaques while presenting surfaces of different depths rendered in dynamic random dot stereograms. We found that the interaction between pairs of neurons was a function of similarity in receptive fields, as well as of the input stimulus. Neurons coding the same depth experienced common inhibition early in their responses for stimuli presented at their nonpreferred disparities. They experienced mutual facilitation later in their responses for stimulation at their preferred disparity. These findings are consistent with a local competition mechanism that first removes gross mismatches, and a global cooperative mechanism that further refines depth estimates. 1 In trod u ction The human visual system is able to extract three-dimensional (3D) structures in random noise stereograms even when such images evoke no perceptible patterns when viewed monocularly [1]. Bela Julesz proposed that this is accomplished by a stereopsis mechanism that detects correlated shifts in 2D noise patterns between the two eyes. He also suggested that this mechanism likely involves cooperative neural processing early in the visual system. Marr and Poggio formalized the computational constraints for solving stereo matching (Fig. 1a) and devised an algorithm that can discover the underlying 3D structures in a variety of random dot stereogram patterns [2]. Their algorithm was based on two rules: (1) each element or feature is unique (i.e., can be assigned only one disparity) and (2) surfaces of objects are cohesive (i.e., depth changes gradually across space). To describe their algorithm in neurophysiological terms, we can consider neurons in primary visual cortex as simple element or feature detectors. The first rule is implemented by introducing competitive interactions (mutual inhibition) among neurons of different disparity tuning at each location (Fig. 1b, blue solid horizontal or vertical lines), allowing only one disparity to be detected at each location. The second rule is implemented by introducing cooperative interactions (mutual facilitation) among neurons tuned to the same depth (image disparity) across different spatial locations (Fig. 1b, along the red dashed diagonal lines). In other words, a disparity estimate at one location is more likely to be correct if neighboring locations have similar disparity estimates. A dynamic system under such constraints can relax to a stable global disparity map. Here, we present neurophysiological evidence of interactions between disparity-tuned neurons in the primary visual cortex that is consistent with this general approach. We sampled from a variety of spatially distributed disparity tuned neurons (see electrodes Fig. 1b) while displaying DRDS stimuli defined at various disparities (see stimulus Fig.1b). We then measured the dynamics of interactions by assessing the temporal evolution of correlation in neural responses. a Left Image b Right Image Electrodes Disparity Left Image ? Stimulus Right Image Figure 1: (a) Left and right images of random dot stereogram (right image has been shifted to the right). (b) 1D graphical depiction of competition (blue solid lines) and cooperation (red dashed lines) among disparity-tuned neurons with respect to space as defined by Marr and Poggio’s stereo algorithm [2]. 2 2.1 Methods Recording and stimulation a Posterior - Anterior Recordings were made in V1 of two awake, behaving macaques. We simultaneously recorded from 4-8 electrodes providing data from up to 10 neurons in a single recording session (some electrodes recorded from as many as 3 neurons). We collected data from 112 neurons that provided 224 pairs for cross-correlation analysis. For stimuli, we used 12 Hz dynamic random dot stereograms (DRDS; 25% density black and white pixels on a mean luminance background) presented in a 3.5-degree aperture. Liquid crystal shutter goggles were used to present random dot patterns to each eye separately. Eleven horizontal disparities between the two eyes, ranging from ±0.9 degrees, were tested. Seventy-four neurons (66%) had significant disparity tuning and 99 pairs (44%) were comprised of neurons that both had significant disparity tuning (1-way ANOVA, p<0.05). b 5mm Medial - Lateral 100µV 0.2ms 1° Figure 2: (a) Example recording session from five electrodes in V1. (b) Receptive field (white box—arrow represents direction preference) and random dot stereogram locations for same recording session (small red square is the fixation spot). 2.2 Data analysis Interaction between neurons was described as
Reference: text
sentIndex sentText sentNum sentScore
1 edu Abstract Although there has been substantial progress in understanding the neurophysiological mechanisms of stereopsis, how neurons interact in a network during stereo computation remains unclear. [sent-9, score-0.505]
2 To test these predictions, we simultaneously recorded from multiple neurons in V1 of awake, behaving macaques while presenting surfaces of different depths rendered in dynamic random dot stereograms. [sent-11, score-0.461]
3 We found that the interaction between pairs of neurons was a function of similarity in receptive fields, as well as of the input stimulus. [sent-12, score-0.464]
4 To describe their algorithm in neurophysiological terms, we can consider neurons in primary visual cortex as simple element or feature detectors. [sent-26, score-0.512]
5 The first rule is implemented by introducing competitive interactions (mutual inhibition) among neurons of different disparity tuning at each location (Fig. [sent-27, score-1.065]
6 1b, blue solid horizontal or vertical lines), allowing only one disparity to be detected at each location. [sent-28, score-0.633]
7 The second rule is implemented by introducing cooperative interactions (mutual facilitation) among neurons tuned to the same depth (image disparity) across different spatial locations (Fig. [sent-29, score-0.54]
8 In other words, a disparity estimate at one location is more likely to be correct if neighboring locations have similar disparity estimates. [sent-31, score-1.137]
9 A dynamic system under such constraints can relax to a stable global disparity map. [sent-32, score-0.557]
10 Here, we present neurophysiological evidence of interactions between disparity-tuned neurons in the primary visual cortex that is consistent with this general approach. [sent-33, score-0.558]
11 We sampled from a variety of spatially distributed disparity tuned neurons (see electrodes Fig. [sent-34, score-0.949]
12 1b) while displaying DRDS stimuli defined at various disparities (see stimulus Fig. [sent-35, score-0.317]
13 We then measured the dynamics of interactions by assessing the temporal evolution of correlation in neural responses. [sent-37, score-0.472]
14 (b) 1D graphical depiction of competition (blue solid lines) and cooperation (red dashed lines) among disparity-tuned neurons with respect to space as defined by Marr and Poggio’s stereo algorithm [2]. [sent-40, score-0.572]
15 We simultaneously recorded from 4-8 electrodes providing data from up to 10 neurons in a single recording session (some electrodes recorded from as many as 3 neurons). [sent-43, score-0.535]
16 We collected data from 112 neurons that provided 224 pairs for cross-correlation analysis. [sent-44, score-0.358]
17 Seventy-four neurons (66%) had significant disparity tuning and 99 pairs (44%) were comprised of neurons that both had significant disparity tuning (1-way ANOVA, p<0. [sent-50, score-2.306]
18 2 Data analysis Interaction between neurons was described as "effective connectivity" defined by crosscorrelation methods [3]. [sent-56, score-0.341]
19 First, the probability of all joint spikes (x and y) between the two neurons was calculated for all times from stimulus onset (t1 and t2) including all possible lag times (t1 - t2) between the two neurons (2D joint peristimulus time histogram—JPSTH). [sent-57, score-0.822]
20 The principal diagonal also represents time from stimulus onset for correlation and the opposite diagonal represents lag time between the two neurons. [sent-60, score-0.54]
21 We first integrated along the principal diagonal to produce correlation versus lag time (i. [sent-63, score-0.441]
22 We used CCHs to find significant correlation at or near 0 ms lag times (suggesting synaptic connectivity between the neurons). [sent-66, score-0.892]
23 Second, we integrated under the half-height full bandwidth of significant correlation peaks to quantify effective connectivity. [sent-67, score-0.611]
24 Finally, we repeated this integration along the principal diagonal to obtain the temporal evolution of effective connectivity (computed with a running 100 ms window). [sent-69, score-0.399]
25 , due to difference in attentional effort in different trials), correlation peaks can emerge that are not due to effective connectivity [5]. [sent-74, score-0.626]
26 We applied a correction to equation 1 [5,6] based on the average firing rate for each trial. [sent-75, score-0.381]
27 However, no significant difference in correlation peaks was observed. [sent-76, score-0.567]
28 In addition, changes in DRDS properties other than disparity did not cause significant changes to correlation peak properties. [sent-77, score-1.089]
29 Finally, alternative cross-correlation methods (CCG) [7] using responses to the same exact random dot pattern to predict correlation expected from chance, again, lead to no significant difference in correlation peak properties. [sent-78, score-0.907]
30 004 -300 Peak Lag Time -100 100 Lag Time (ms) 300 Figure 4: (a) Population average CCH for 27 neuron pairs with a significant correlation peak. [sent-91, score-0.61]
31 (b) Same as (a), but zoomed into ±50 ms lag times with statistics of peak properties (mean ± s. [sent-92, score-0.352]
32 3 Interaction depends on tuning properties The primary indicator of whether or not a neuron pair had a significant correlation peak at or near a 0 ms lag time, for this class of stimuli, was similarity in disparity tuning between the two neurons. [sent-96, score-1.807]
33 Neuron pairs with significant correlation peaks (n = 27; 27%) tended to have more similar disparity peaks, bandwidths, and frequencies (determined from fitted Gabor functions) than neuron pairs that did not have significant correlation peaks. [sent-97, score-1.792]
34 4 Correlation Percentage Pairs ( n=27 ) where i is each point on the disparity tuning curve, x and y are the firing rates at each point for each neuron, and x and y are the mean firing rates across the tuning curve. [sent-103, score-1.627]
35 Figure 5a and 5b clearly show that both the probability of correlation and strength in correlation increase with greater SI (n = 27 pairs). [sent-104, score-0.636]
36 This relationship is limited to long-range interactions among neurons because our electrodes were all at least 1 mm apart. [sent-105, score-0.433]
37 This suggests they are likely mediated by the well known long-range intracortical connections in V1 that link neurons of similar orientation across space [9]. [sent-106, score-0.474]
38 Our results suggest that these connections might also be shared to link similar disparity neurons together. [sent-107, score-0.898]
39 Because connectivity also depended on orientation (Figure 5c), V1 connectivity among neurons appears to depend on similarity across multiple cue dimensions. [sent-108, score-0.679]
40 0 0 20 40 60 80 ∆ Orientation Preference Figure 5: (a) Likelihood of significant correlation peak with respect to similar disparity tuning. [sent-126, score-1.089]
41 From the 12 pairs of neurons recorded on a single electrode, correlation was observed among neuron pairs with very similar disparity tuning as well as among neurons with nearly opposite disparity tuning (see also [8]). [sent-129, score-2.601]
42 This suggests that antagonistic disparity-tuned neurons tend to spatially coexist, and their interactions are likely competitive. [sent-130, score-0.477]
43 4 Interaction is stimulus-dependent The interaction between pairs of neurons was not simply a function of the similarity between their receptive field properties but was also a function of the input stimuli (or stimulus disparity in our case). [sent-131, score-1.124]
44 05) by the stimulus disparity for 25 out of the 27 pairs. [sent-133, score-0.622]
45 For neuron pairs with similar disparity tuning, the strongest correlation was observed at their shared preferred disparity, i. [sent-135, score-1.171]
46 the peak of the disparity tuning curves based on firing rate (as shown in Figure 6). [sent-137, score-1.159]
47 This suggests facilitation is strongest when a frontal parallel plane activated these neurons simultaneously at their preferred depth. [sent-138, score-0.602]
48 As the stimulus plane moved away from this depth, the effective connectivity between the neurons became weaker. [sent-139, score-0.554]
49 , Figure 6d), the correlation or effective connectivity was again strongest at the neuron pair's shared preferred disparity. [sent-145, score-0.745]
50 However, these pairs in addition exhibited secondary correlation peaks for disparity stimuli that produced the lowest firing rates (even below the baseline for DRDSs). [sent-146, score-1.488]
51 50 sps 80 sps b 40 sps 20 sps Firing Rate a -0. [sent-147, score-0.312]
52 9 Horizontal Disparity (°) Horizontal Disparity (°) Figure 6: Top row are disparity tuning curves based on firing rates (mean ± s. [sent-167, score-1.092]
53 Bottom row are disparity tuning curves based on correlation for the corresponding pairs of neurons in the top row. [sent-171, score-1.37]
54 As correlation peaks become broader, as in our case (mean = 42 ms), this interpretation becomes more ambiguous (more possible circuits). [sent-174, score-0.437]
55 The broader positive correlation peaks can even be caused by common inhibitory circuitry. [sent-175, score-0.437]
56 One way to potentially disambiguate our interpretations is to consider firing rate behavior. [sent-176, score-0.381]
57 The positive correlation measured at the preferred disparity suggests that the interaction was likely facilitatory in nature based on the increased firing of the neurons. [sent-177, score-1.448]
58 The positive correlation measured at the disparity where both neurons' firing rates were depressed, i. [sent-178, score-1.273]
59 at the valley of the firing- rate based disparity tuning curves, suggests that the correlation likely arose from common inhibition (presumably from neurons that preferred that disparity). [sent-180, score-1.576]
60 5 Temporal dynamics of interaction We can compare the temporal dynamics of the correlation with the temporal dynamics of the firing rate of the neurons to gain more insight into the possible underlying circuitry. [sent-181, score-1.252]
61 We computed the correlation every 1 msec over a 100 ms running window, and found that the correlation peak at the preferred disparity (based on firing rates) occurred at a later time (250-350 ms post-stimulus onset) than the correlation peaks at the non-preferred disparity (100-200 ms). [sent-182, score-3.056]
62 Figure 7 illustrates the temporal dynamics of correlation for the example neuron pair shown in Figure 6b and 6d. [sent-183, score-0.502]
63 The distinct interval in which correlation emerged at the preferred and the non-preferred disparities was consistently observed for all 27 pairs of neurons. [sent-184, score-0.642]
64 Even for the example shown in Figure 6c, there were peaks in correlation in the early part of the response at the most non-preferred disparities. [sent-185, score-0.469]
65 The timing of these two phases of correlation was also rather consistent over the population of pairs. [sent-186, score-0.354]
66 1 Horizontal Disparity (°) Figure 7: Temporal dynamics of correlation for example neuron pair shown in Figure 6, right. [sent-202, score-0.465]
67 Disparity tuning based on correlation for the early (blue) and late (red) portion of the response (95% confidence intervals). [sent-205, score-0.565]
68 Correlation was calculated every 1 ms over 100 ms windows. [sent-206, score-0.29]
69 By examining the interplay between firing rate and correlation, we were able to gain even greater insight about the interactions among neuron pairs. [sent-207, score-0.581]
70 To summarize this interplay across our population, we compared the temporal evolution of the correlation at three distinct disparities with the temporal evolution of the firing rates at the same disparities (also smoothed with 100 ms time windows). [sent-208, score-1.362]
71 The first disparity, the preferred disparity A, is where we measured the strongest correlation and was at or near the highest firing rate measured in individual neurons (see Figure 8, left). [sent-209, score-1.69]
72 The second important disparity, the most non-preferred disparity C, was where we measured secondary correlation peaks and coincided with the lowest firing rates observed in individual neurons. [sent-210, score-1.392]
73 Lastly, we looked at a disparity B that was in between disparities A and C. [sent-211, score-0.73]
74 Figure 8 shows that neurons responded better to their preferred disparity over other disparities very early, resulting in immediate moderate firing rate-based disparity tuning. [sent-212, score-2.036]
75 Then shortly after (100 ms), a correlation peak emerges at the least preferred disparity C. [sent-213, score-1.052]
76 This coincides with suppression of firing rate for all disparities (Figure 8, blue dashed line). [sent-214, score-0.608]
77 However, the suppression in firing rate is much stronger for C where the firing rate diverges downward from the firing rates for A and B sharpening the disparity tuning (Figure 8, blue arrow; see also [11]). [sent-215, score-1.885]
78 The strong correlation coupled with the decrease in firing suggests strong common inhibition. [sent-216, score-0.725]
79 05 30 0 100 A B C 25 20 200 300 400 300 400 n = 32 cells 15 10 5 0 -100 0 100 200 Time (ms) Figure 8: Population average of normalized correlation versus time (top) for three disparities shown on the left. [sent-224, score-0.491]
80 Both correlation and firing rates were calculated every 1 ms over 100 ms windows. [sent-226, score-1.006]
81 Once the correlation peak at C subsided (200 ms), the correlation increased for A (red dashed line). [sent-227, score-0.743]
82 When the correlation for A peaked, the correlation decreased for B and C, leading to very sharp correlation-based disparity tuning (see also Figure 7). [sent-228, score-1.33]
83 This correlationbased tuning can facilitate depth estimates by changing how effectively these signals are integrated downstream as a function of disparity [12]. [sent-229, score-0.77]
84 Our interpretation is that the initial firing rate bias leads to antagonistic disparity-tuned neurons generating common inhibition that suppresses firing at non-preferred disparities, removing potential mismatches. [sent-230, score-1.136]
85 The immediacy suggests that mutual inhibition was local, which is consistent with our observation that many opposing disparity-tuned neurons spatially coexisted. [sent-231, score-0.483]
86 The slower correlation peak at the preferred disparity A is indicative of mutual facilitation that occurred when the depth estimates of spatially distinct neurons matched. [sent-232, score-1.629]
87 More recent cooperative stereo computation models allow excitatory interaction between neurons of different disparities separated by long distance. [sent-235, score-0.71]
88 In this experiment, we only tested frontal parallel planes, thus, we cannot answer whether or not effective connections and facilitation exist between neurons with larger disparity differences over long distances. [sent-237, score-1.059]
89 This will require further experiments using planes with disparity gradients. [sent-238, score-0.58]
90 The observation that initial correlation peaks occurred at disparities that evoked the lowest firing rates in neurons, suggests that correlation peaks emerged from common inhibition for non-preferred disparities. [sent-239, score-1.637]
91 The observation that later correlation occurred at disparities that evoked the highest firing rates suggests that neurons were mutually exciting each other at their preferred disparity. [sent-240, score-1.402]
92 Our neurophysiological data reveal interesting dynamics between network-based (effective connectivity) and firing rate-based encoding of depth estimates. [sent-241, score-0.555]
93 The observation that inhibition precedes facilitation suggests that competition is local (re- calling neurons at the same electrode tend to have opposite disparity tuning) and cooperation is more global (mediated through long-range connectivity). [sent-242, score-1.164]
94 Local competition between neurons encoding different depths is consistent with the uniqueness principle of Marr and Poggio's algorithm [2]. [sent-243, score-0.366]
95 In addition, cooperation among neurons encoding the same depth across space was predicted by the second rule of their algorithm: matter is cohesive. [sent-244, score-0.447]
96 Previous neurophysiological data had suggested that intracortical connectivity in primary visual cortex underlies competitive [15] and cooperative [16] mechanisms for improving estimates of orientation. [sent-246, score-0.506]
97 (1989) Dynamics of neuronal firing correlation: modulation of "effective connectivity". [sent-266, score-0.356]
98 (2005) Stimulus dependence of neuronal correlation in primary visual cortex of the macaque. [sent-285, score-0.45]
99 (2004) Functional connectivity of disparity-tuned neurons in the visual cortex. [sent-290, score-0.49]
100 (1985) PMF: a stereo correspondence algorithm using a disparity gradient limit. [sent-326, score-0.651]
wordName wordTfidf (topN-words)
[('disparity', 0.557), ('firing', 0.356), ('correlation', 0.318), ('neurons', 0.3), ('disparities', 0.173), ('connectivity', 0.145), ('ms', 0.145), ('tuning', 0.137), ('significant', 0.13), ('lag', 0.123), ('peaks', 0.119), ('neuron', 0.104), ('facilitation', 0.095), ('stereo', 0.094), ('preferred', 0.093), ('cooperative', 0.093), ('peak', 0.084), ('neurophysiological', 0.08), ('confidence', 0.078), ('marr', 0.078), ('sps', 0.078), ('depth', 0.076), ('inhibition', 0.072), ('stimulus', 0.065), ('jpsth', 0.062), ('electrodes', 0.062), ('pairs', 0.058), ('dot', 0.057), ('poggio', 0.054), ('suggests', 0.051), ('interaction', 0.05), ('cch', 0.047), ('cnbc', 0.047), ('drds', 0.047), ('julesz', 0.047), ('stereogram', 0.047), ('interactions', 0.046), ('cooperation', 0.046), ('occurred', 0.046), ('visual', 0.045), ('horizontal', 0.045), ('cortex', 0.045), ('effective', 0.044), ('dynamics', 0.043), ('competition', 0.043), ('rates', 0.042), ('primary', 0.042), ('strongest', 0.041), ('connections', 0.041), ('defined', 0.041), ('gerstein', 0.041), ('stereograms', 0.041), ('stereopsis', 0.041), ('stimuli', 0.038), ('temporal', 0.037), ('population', 0.036), ('orientation', 0.034), ('onset', 0.034), ('recording', 0.033), ('early', 0.032), ('mechanisms', 0.031), ('synaptic', 0.031), ('assemblies', 0.031), ('cchs', 0.031), ('menz', 0.031), ('samonds', 0.031), ('surfaces', 0.031), ('blue', 0.031), ('spatially', 0.03), ('similarity', 0.03), ('mutual', 0.03), ('session', 0.028), ('evolution', 0.028), ('tai', 0.027), ('antagonistic', 0.027), ('red', 0.027), ('histogram', 0.026), ('receptive', 0.026), ('recorded', 0.025), ('rate', 0.025), ('among', 0.025), ('carnegie', 0.025), ('mellon', 0.025), ('pittsburgh', 0.025), ('findings', 0.025), ('interplay', 0.025), ('intracortical', 0.025), ('awake', 0.025), ('behaving', 0.025), ('si', 0.025), ('preference', 0.023), ('dashed', 0.023), ('evoked', 0.023), ('planes', 0.023), ('anova', 0.023), ('experienced', 0.023), ('depths', 0.023), ('likely', 0.023), ('frontal', 0.022), ('arrow', 0.022)]
simIndex simValue paperId paperTitle
same-paper 1 1.0000004 145 nips-2006-Neurophysiological Evidence of Cooperative Mechanisms for Stereo Computation
Author: Jason M. Samonds, Brian R. Potetz, Tai S. Lee
Abstract: Although there has been substantial progress in understanding the neurophysiological mechanisms of stereopsis, how neurons interact in a network during stereo computation remains unclear. Computational models on stereopsis suggest local competition and long-range cooperation are important for resolving ambiguity during stereo matching. To test these predictions, we simultaneously recorded from multiple neurons in V1 of awake, behaving macaques while presenting surfaces of different depths rendered in dynamic random dot stereograms. We found that the interaction between pairs of neurons was a function of similarity in receptive fields, as well as of the input stimulus. Neurons coding the same depth experienced common inhibition early in their responses for stimuli presented at their nonpreferred disparities. They experienced mutual facilitation later in their responses for stimulation at their preferred disparity. These findings are consistent with a local competition mechanism that first removes gross mismatches, and a global cooperative mechanism that further refines depth estimates. 1 In trod u ction The human visual system is able to extract three-dimensional (3D) structures in random noise stereograms even when such images evoke no perceptible patterns when viewed monocularly [1]. Bela Julesz proposed that this is accomplished by a stereopsis mechanism that detects correlated shifts in 2D noise patterns between the two eyes. He also suggested that this mechanism likely involves cooperative neural processing early in the visual system. Marr and Poggio formalized the computational constraints for solving stereo matching (Fig. 1a) and devised an algorithm that can discover the underlying 3D structures in a variety of random dot stereogram patterns [2]. Their algorithm was based on two rules: (1) each element or feature is unique (i.e., can be assigned only one disparity) and (2) surfaces of objects are cohesive (i.e., depth changes gradually across space). To describe their algorithm in neurophysiological terms, we can consider neurons in primary visual cortex as simple element or feature detectors. The first rule is implemented by introducing competitive interactions (mutual inhibition) among neurons of different disparity tuning at each location (Fig. 1b, blue solid horizontal or vertical lines), allowing only one disparity to be detected at each location. The second rule is implemented by introducing cooperative interactions (mutual facilitation) among neurons tuned to the same depth (image disparity) across different spatial locations (Fig. 1b, along the red dashed diagonal lines). In other words, a disparity estimate at one location is more likely to be correct if neighboring locations have similar disparity estimates. A dynamic system under such constraints can relax to a stable global disparity map. Here, we present neurophysiological evidence of interactions between disparity-tuned neurons in the primary visual cortex that is consistent with this general approach. We sampled from a variety of spatially distributed disparity tuned neurons (see electrodes Fig. 1b) while displaying DRDS stimuli defined at various disparities (see stimulus Fig.1b). We then measured the dynamics of interactions by assessing the temporal evolution of correlation in neural responses. a Left Image b Right Image Electrodes Disparity Left Image ? Stimulus Right Image Figure 1: (a) Left and right images of random dot stereogram (right image has been shifted to the right). (b) 1D graphical depiction of competition (blue solid lines) and cooperation (red dashed lines) among disparity-tuned neurons with respect to space as defined by Marr and Poggio’s stereo algorithm [2]. 2 2.1 Methods Recording and stimulation a Posterior - Anterior Recordings were made in V1 of two awake, behaving macaques. We simultaneously recorded from 4-8 electrodes providing data from up to 10 neurons in a single recording session (some electrodes recorded from as many as 3 neurons). We collected data from 112 neurons that provided 224 pairs for cross-correlation analysis. For stimuli, we used 12 Hz dynamic random dot stereograms (DRDS; 25% density black and white pixels on a mean luminance background) presented in a 3.5-degree aperture. Liquid crystal shutter goggles were used to present random dot patterns to each eye separately. Eleven horizontal disparities between the two eyes, ranging from ±0.9 degrees, were tested. Seventy-four neurons (66%) had significant disparity tuning and 99 pairs (44%) were comprised of neurons that both had significant disparity tuning (1-way ANOVA, p<0.05). b 5mm Medial - Lateral 100µV 0.2ms 1° Figure 2: (a) Example recording session from five electrodes in V1. (b) Receptive field (white box—arrow represents direction preference) and random dot stereogram locations for same recording session (small red square is the fixation spot). 2.2 Data analysis Interaction between neurons was described as
2 0.30447379 15 nips-2006-A Switched Gaussian Process for Estimating Disparity and Segmentation in Binocular Stereo
Author: Oliver Williams
Abstract: This paper describes a Gaussian process framework for inferring pixel-wise disparity and bi-layer segmentation of a scene given a stereo pair of images. The Gaussian process covariance is parameterized by a foreground-backgroundocclusion segmentation label to model both smooth regions and discontinuities. As such, we call our model a switched Gaussian process. We propose a greedy incremental algorithm for adding observations from the data and assigning segmentation labels. Two observation schedules are proposed: the first treats scanlines as independent, the second uses an active learning criterion to select a sparse subset of points to measure. We show that this probabilistic framework has comparable performance to the state-of-the-art. 1
3 0.28928977 59 nips-2006-Context dependent amplification of both rate and event-correlation in a VLSI network of spiking neurons
Author: Elisabetta Chicca, Giacomo Indiveri, Rodney J. Douglas
Abstract: Cooperative competitive networks are believed to play a central role in cortical processing and have been shown to exhibit a wide set of useful computational properties. We propose a VLSI implementation of a spiking cooperative competitive network and show how it can perform context dependent computation both in the mean firing rate domain and in spike timing correlation space. In the mean rate case the network amplifies the activity of neurons belonging to the selected stimulus and suppresses the activity of neurons receiving weaker stimuli. In the event correlation case, the recurrent network amplifies with a higher gain the correlation between neurons which receive highly correlated inputs while leaving the mean firing rate unaltered. We describe the network architecture and present experimental data demonstrating its context dependent computation capabilities. 1
4 0.20049725 36 nips-2006-Attentional Processing on a Spike-Based VLSI Neural Network
Author: Yingxue Wang, Rodney J. Douglas, Shih-Chii Liu
Abstract: The neurons of the neocortex communicate by asynchronous events called action potentials (or ’spikes’). However, for simplicity of simulation, most models of processing by cortical neural networks have assumed that the activations of their neurons can be approximated by event rates rather than taking account of individual spikes. The obstacle to exploring the more detailed spike processing of these networks has been reduced considerably in recent years by the development of hybrid analog-digital Very-Large Scale Integrated (hVLSI) neural networks composed of spiking neurons that are able to operate in real-time. In this paper we describe such a hVLSI neural network that performs an interesting task of selective attentional processing that was previously described for a simulated ’pointer-map’ rate model by Hahnloser and colleagues. We found that most of the computational features of their rate model can be reproduced in the spiking implementation; but, that spike-based processing requires a modification of the original network architecture in order to memorize a previously attended target. 1
5 0.18884088 197 nips-2006-Uncertainty, phase and oscillatory hippocampal recall
Author: Máté Lengyel, Peter Dayan
Abstract: Many neural areas, notably, the hippocampus, show structured, dynamical, population behavior such as coordinated oscillations. It has long been observed that such oscillations provide a substrate for representing analog information in the firing phases of neurons relative to the underlying population rhythm. However, it has become increasingly clear that it is essential for neural populations to represent uncertainty about the information they capture, and the substantial recent work on neural codes for uncertainty has omitted any analysis of oscillatory systems. Here, we observe that, since neurons in an oscillatory network need not only fire once in each cycle (or even at all), uncertainty about the analog quantities each neuron represents by its firing phase might naturally be reported through the degree of concentration of the spikes that it fires. We apply this theory to memory in a model of oscillatory associative recall in hippocampal area CA3. Although it is not well treated in the literature, representing and manipulating uncertainty is fundamental to competent memory; our theory enables us to view CA3 as an effective uncertainty-aware, retrieval system. 1
6 0.1770989 187 nips-2006-Temporal Coding using the Response Properties of Spiking Neurons
7 0.17494629 189 nips-2006-Temporal dynamics of information content carried by neurons in the primary visual cortex
8 0.1296906 99 nips-2006-Information Bottleneck Optimization and Independent Component Extraction with Spiking Neurons
9 0.12472477 154 nips-2006-Optimal Change-Detection and Spiking Neurons
10 0.10321876 113 nips-2006-Learning Structural Equation Models for fMRI
11 0.07301221 18 nips-2006-A selective attention multi--chip system with dynamic synapses and spiking neurons
12 0.068844154 16 nips-2006-A Theory of Retinal Population Coding
13 0.05478818 162 nips-2006-Predicting spike times from subthreshold dynamics of a neuron
14 0.053434923 165 nips-2006-Real-time adaptive information-theoretic optimization of neurophysiology experiments
15 0.051256489 8 nips-2006-A Nonparametric Approach to Bottom-Up Visual Saliency
16 0.046499789 148 nips-2006-Nonlinear physically-based models for decoding motor-cortical population activity
17 0.043325767 116 nips-2006-Learning from Multiple Sources
18 0.04082071 75 nips-2006-Efficient sparse coding algorithms
19 0.038828775 12 nips-2006-A Probabilistic Algorithm Integrating Source Localization and Noise Suppression of MEG and EEG data
20 0.038627271 22 nips-2006-Adaptive Spatial Filters with predefined Region of Interest for EEG based Brain-Computer-Interfaces
topicId topicWeight
[(0, -0.142), (1, -0.353), (2, 0.068), (3, 0.068), (4, 0.024), (5, 0.015), (6, -0.004), (7, 0.054), (8, -0.002), (9, -0.007), (10, 0.1), (11, -0.001), (12, -0.05), (13, -0.032), (14, -0.086), (15, 0.073), (16, 0.106), (17, -0.056), (18, -0.046), (19, 0.18), (20, -0.093), (21, 0.114), (22, 0.103), (23, 0.027), (24, 0.154), (25, -0.228), (26, -0.067), (27, -0.201), (28, 0.103), (29, 0.016), (30, 0.019), (31, -0.031), (32, -0.008), (33, 0.02), (34, -0.223), (35, -0.059), (36, 0.032), (37, 0.075), (38, -0.237), (39, 0.077), (40, 0.017), (41, -0.024), (42, 0.025), (43, -0.051), (44, -0.016), (45, -0.15), (46, 0.003), (47, 0.094), (48, -0.062), (49, 0.077)]
simIndex simValue paperId paperTitle
same-paper 1 0.9859131 145 nips-2006-Neurophysiological Evidence of Cooperative Mechanisms for Stereo Computation
Author: Jason M. Samonds, Brian R. Potetz, Tai S. Lee
Abstract: Although there has been substantial progress in understanding the neurophysiological mechanisms of stereopsis, how neurons interact in a network during stereo computation remains unclear. Computational models on stereopsis suggest local competition and long-range cooperation are important for resolving ambiguity during stereo matching. To test these predictions, we simultaneously recorded from multiple neurons in V1 of awake, behaving macaques while presenting surfaces of different depths rendered in dynamic random dot stereograms. We found that the interaction between pairs of neurons was a function of similarity in receptive fields, as well as of the input stimulus. Neurons coding the same depth experienced common inhibition early in their responses for stimuli presented at their nonpreferred disparities. They experienced mutual facilitation later in their responses for stimulation at their preferred disparity. These findings are consistent with a local competition mechanism that first removes gross mismatches, and a global cooperative mechanism that further refines depth estimates. 1 In trod u ction The human visual system is able to extract three-dimensional (3D) structures in random noise stereograms even when such images evoke no perceptible patterns when viewed monocularly [1]. Bela Julesz proposed that this is accomplished by a stereopsis mechanism that detects correlated shifts in 2D noise patterns between the two eyes. He also suggested that this mechanism likely involves cooperative neural processing early in the visual system. Marr and Poggio formalized the computational constraints for solving stereo matching (Fig. 1a) and devised an algorithm that can discover the underlying 3D structures in a variety of random dot stereogram patterns [2]. Their algorithm was based on two rules: (1) each element or feature is unique (i.e., can be assigned only one disparity) and (2) surfaces of objects are cohesive (i.e., depth changes gradually across space). To describe their algorithm in neurophysiological terms, we can consider neurons in primary visual cortex as simple element or feature detectors. The first rule is implemented by introducing competitive interactions (mutual inhibition) among neurons of different disparity tuning at each location (Fig. 1b, blue solid horizontal or vertical lines), allowing only one disparity to be detected at each location. The second rule is implemented by introducing cooperative interactions (mutual facilitation) among neurons tuned to the same depth (image disparity) across different spatial locations (Fig. 1b, along the red dashed diagonal lines). In other words, a disparity estimate at one location is more likely to be correct if neighboring locations have similar disparity estimates. A dynamic system under such constraints can relax to a stable global disparity map. Here, we present neurophysiological evidence of interactions between disparity-tuned neurons in the primary visual cortex that is consistent with this general approach. We sampled from a variety of spatially distributed disparity tuned neurons (see electrodes Fig. 1b) while displaying DRDS stimuli defined at various disparities (see stimulus Fig.1b). We then measured the dynamics of interactions by assessing the temporal evolution of correlation in neural responses. a Left Image b Right Image Electrodes Disparity Left Image ? Stimulus Right Image Figure 1: (a) Left and right images of random dot stereogram (right image has been shifted to the right). (b) 1D graphical depiction of competition (blue solid lines) and cooperation (red dashed lines) among disparity-tuned neurons with respect to space as defined by Marr and Poggio’s stereo algorithm [2]. 2 2.1 Methods Recording and stimulation a Posterior - Anterior Recordings were made in V1 of two awake, behaving macaques. We simultaneously recorded from 4-8 electrodes providing data from up to 10 neurons in a single recording session (some electrodes recorded from as many as 3 neurons). We collected data from 112 neurons that provided 224 pairs for cross-correlation analysis. For stimuli, we used 12 Hz dynamic random dot stereograms (DRDS; 25% density black and white pixels on a mean luminance background) presented in a 3.5-degree aperture. Liquid crystal shutter goggles were used to present random dot patterns to each eye separately. Eleven horizontal disparities between the two eyes, ranging from ±0.9 degrees, were tested. Seventy-four neurons (66%) had significant disparity tuning and 99 pairs (44%) were comprised of neurons that both had significant disparity tuning (1-way ANOVA, p<0.05). b 5mm Medial - Lateral 100µV 0.2ms 1° Figure 2: (a) Example recording session from five electrodes in V1. (b) Receptive field (white box—arrow represents direction preference) and random dot stereogram locations for same recording session (small red square is the fixation spot). 2.2 Data analysis Interaction between neurons was described as
2 0.64901048 15 nips-2006-A Switched Gaussian Process for Estimating Disparity and Segmentation in Binocular Stereo
Author: Oliver Williams
Abstract: This paper describes a Gaussian process framework for inferring pixel-wise disparity and bi-layer segmentation of a scene given a stereo pair of images. The Gaussian process covariance is parameterized by a foreground-backgroundocclusion segmentation label to model both smooth regions and discontinuities. As such, we call our model a switched Gaussian process. We propose a greedy incremental algorithm for adding observations from the data and assigning segmentation labels. Two observation schedules are proposed: the first treats scanlines as independent, the second uses an active learning criterion to select a sparse subset of points to measure. We show that this probabilistic framework has comparable performance to the state-of-the-art. 1
3 0.60370076 59 nips-2006-Context dependent amplification of both rate and event-correlation in a VLSI network of spiking neurons
Author: Elisabetta Chicca, Giacomo Indiveri, Rodney J. Douglas
Abstract: Cooperative competitive networks are believed to play a central role in cortical processing and have been shown to exhibit a wide set of useful computational properties. We propose a VLSI implementation of a spiking cooperative competitive network and show how it can perform context dependent computation both in the mean firing rate domain and in spike timing correlation space. In the mean rate case the network amplifies the activity of neurons belonging to the selected stimulus and suppresses the activity of neurons receiving weaker stimuli. In the event correlation case, the recurrent network amplifies with a higher gain the correlation between neurons which receive highly correlated inputs while leaving the mean firing rate unaltered. We describe the network architecture and present experimental data demonstrating its context dependent computation capabilities. 1
4 0.56321937 189 nips-2006-Temporal dynamics of information content carried by neurons in the primary visual cortex
Author: Danko Nikolić, Stefan Haeusler, Wolf Singer, Wolfgang Maass
Abstract: We use multi-electrode recordings from cat primary visual cortex and investigate whether a simple linear classifier can extract information about the presented stimuli. We find that information is extractable and that it even lasts for several hundred milliseconds after the stimulus has been removed. In a fast sequence of stimulus presentation, information about both new and old stimuli is present simultaneously and nonlinear relations between these stimuli can be extracted. These results suggest nonlinear properties of cortical representations. The important implications of these properties for the nonlinear brain theory are discussed.
5 0.55314606 36 nips-2006-Attentional Processing on a Spike-Based VLSI Neural Network
Author: Yingxue Wang, Rodney J. Douglas, Shih-Chii Liu
Abstract: The neurons of the neocortex communicate by asynchronous events called action potentials (or ’spikes’). However, for simplicity of simulation, most models of processing by cortical neural networks have assumed that the activations of their neurons can be approximated by event rates rather than taking account of individual spikes. The obstacle to exploring the more detailed spike processing of these networks has been reduced considerably in recent years by the development of hybrid analog-digital Very-Large Scale Integrated (hVLSI) neural networks composed of spiking neurons that are able to operate in real-time. In this paper we describe such a hVLSI neural network that performs an interesting task of selective attentional processing that was previously described for a simulated ’pointer-map’ rate model by Hahnloser and colleagues. We found that most of the computational features of their rate model can be reproduced in the spiking implementation; but, that spike-based processing requires a modification of the original network architecture in order to memorize a previously attended target. 1
6 0.48234412 197 nips-2006-Uncertainty, phase and oscillatory hippocampal recall
7 0.3814916 18 nips-2006-A selective attention multi--chip system with dynamic synapses and spiking neurons
8 0.36295858 187 nips-2006-Temporal Coding using the Response Properties of Spiking Neurons
9 0.31152785 113 nips-2006-Learning Structural Equation Models for fMRI
10 0.26147383 154 nips-2006-Optimal Change-Detection and Spiking Neurons
11 0.25142914 153 nips-2006-Online Clustering of Moving Hyperplanes
12 0.23080637 99 nips-2006-Information Bottleneck Optimization and Independent Component Extraction with Spiking Neurons
13 0.2122618 174 nips-2006-Similarity by Composition
14 0.19665509 192 nips-2006-Theory and Dynamics of Perceptual Bistability
15 0.17754126 51 nips-2006-Clustering Under Prior Knowledge with Application to Image Segmentation
16 0.16637284 190 nips-2006-The Neurodynamics of Belief Propagation on Binary Markov Random Fields
17 0.14801151 98 nips-2006-Inferring Network Structure from Co-Occurrences
18 0.14704669 49 nips-2006-Causal inference in sensorimotor integration
19 0.14388953 1 nips-2006-A Bayesian Approach to Diffusion Models of Decision-Making and Response Time
20 0.14052296 22 nips-2006-Adaptive Spatial Filters with predefined Region of Interest for EEG based Brain-Computer-Interfaces
topicId topicWeight
[(1, 0.05), (3, 0.014), (6, 0.013), (7, 0.05), (9, 0.045), (20, 0.022), (22, 0.039), (30, 0.011), (44, 0.033), (57, 0.036), (65, 0.036), (69, 0.017), (71, 0.492), (90, 0.015), (93, 0.024), (98, 0.01)]
simIndex simValue paperId paperTitle
same-paper 1 0.94155341 145 nips-2006-Neurophysiological Evidence of Cooperative Mechanisms for Stereo Computation
Author: Jason M. Samonds, Brian R. Potetz, Tai S. Lee
Abstract: Although there has been substantial progress in understanding the neurophysiological mechanisms of stereopsis, how neurons interact in a network during stereo computation remains unclear. Computational models on stereopsis suggest local competition and long-range cooperation are important for resolving ambiguity during stereo matching. To test these predictions, we simultaneously recorded from multiple neurons in V1 of awake, behaving macaques while presenting surfaces of different depths rendered in dynamic random dot stereograms. We found that the interaction between pairs of neurons was a function of similarity in receptive fields, as well as of the input stimulus. Neurons coding the same depth experienced common inhibition early in their responses for stimuli presented at their nonpreferred disparities. They experienced mutual facilitation later in their responses for stimulation at their preferred disparity. These findings are consistent with a local competition mechanism that first removes gross mismatches, and a global cooperative mechanism that further refines depth estimates. 1 In trod u ction The human visual system is able to extract three-dimensional (3D) structures in random noise stereograms even when such images evoke no perceptible patterns when viewed monocularly [1]. Bela Julesz proposed that this is accomplished by a stereopsis mechanism that detects correlated shifts in 2D noise patterns between the two eyes. He also suggested that this mechanism likely involves cooperative neural processing early in the visual system. Marr and Poggio formalized the computational constraints for solving stereo matching (Fig. 1a) and devised an algorithm that can discover the underlying 3D structures in a variety of random dot stereogram patterns [2]. Their algorithm was based on two rules: (1) each element or feature is unique (i.e., can be assigned only one disparity) and (2) surfaces of objects are cohesive (i.e., depth changes gradually across space). To describe their algorithm in neurophysiological terms, we can consider neurons in primary visual cortex as simple element or feature detectors. The first rule is implemented by introducing competitive interactions (mutual inhibition) among neurons of different disparity tuning at each location (Fig. 1b, blue solid horizontal or vertical lines), allowing only one disparity to be detected at each location. The second rule is implemented by introducing cooperative interactions (mutual facilitation) among neurons tuned to the same depth (image disparity) across different spatial locations (Fig. 1b, along the red dashed diagonal lines). In other words, a disparity estimate at one location is more likely to be correct if neighboring locations have similar disparity estimates. A dynamic system under such constraints can relax to a stable global disparity map. Here, we present neurophysiological evidence of interactions between disparity-tuned neurons in the primary visual cortex that is consistent with this general approach. We sampled from a variety of spatially distributed disparity tuned neurons (see electrodes Fig. 1b) while displaying DRDS stimuli defined at various disparities (see stimulus Fig.1b). We then measured the dynamics of interactions by assessing the temporal evolution of correlation in neural responses. a Left Image b Right Image Electrodes Disparity Left Image ? Stimulus Right Image Figure 1: (a) Left and right images of random dot stereogram (right image has been shifted to the right). (b) 1D graphical depiction of competition (blue solid lines) and cooperation (red dashed lines) among disparity-tuned neurons with respect to space as defined by Marr and Poggio’s stereo algorithm [2]. 2 2.1 Methods Recording and stimulation a Posterior - Anterior Recordings were made in V1 of two awake, behaving macaques. We simultaneously recorded from 4-8 electrodes providing data from up to 10 neurons in a single recording session (some electrodes recorded from as many as 3 neurons). We collected data from 112 neurons that provided 224 pairs for cross-correlation analysis. For stimuli, we used 12 Hz dynamic random dot stereograms (DRDS; 25% density black and white pixels on a mean luminance background) presented in a 3.5-degree aperture. Liquid crystal shutter goggles were used to present random dot patterns to each eye separately. Eleven horizontal disparities between the two eyes, ranging from ±0.9 degrees, were tested. Seventy-four neurons (66%) had significant disparity tuning and 99 pairs (44%) were comprised of neurons that both had significant disparity tuning (1-way ANOVA, p<0.05). b 5mm Medial - Lateral 100µV 0.2ms 1° Figure 2: (a) Example recording session from five electrodes in V1. (b) Receptive field (white box—arrow represents direction preference) and random dot stereogram locations for same recording session (small red square is the fixation spot). 2.2 Data analysis Interaction between neurons was described as
2 0.90798277 36 nips-2006-Attentional Processing on a Spike-Based VLSI Neural Network
Author: Yingxue Wang, Rodney J. Douglas, Shih-Chii Liu
Abstract: The neurons of the neocortex communicate by asynchronous events called action potentials (or ’spikes’). However, for simplicity of simulation, most models of processing by cortical neural networks have assumed that the activations of their neurons can be approximated by event rates rather than taking account of individual spikes. The obstacle to exploring the more detailed spike processing of these networks has been reduced considerably in recent years by the development of hybrid analog-digital Very-Large Scale Integrated (hVLSI) neural networks composed of spiking neurons that are able to operate in real-time. In this paper we describe such a hVLSI neural network that performs an interesting task of selective attentional processing that was previously described for a simulated ’pointer-map’ rate model by Hahnloser and colleagues. We found that most of the computational features of their rate model can be reproduced in the spiking implementation; but, that spike-based processing requires a modification of the original network architecture in order to memorize a previously attended target. 1
3 0.79491705 135 nips-2006-Modelling transcriptional regulation using Gaussian Processes
Author: Neil D. Lawrence, Guido Sanguinetti, Magnus Rattray
Abstract: Modelling the dynamics of transcriptional processes in the cell requires the knowledge of a number of key biological quantities. While some of them are relatively easy to measure, such as mRNA decay rates and mRNA abundance levels, it is still very hard to measure the active concentration levels of the transcription factor proteins that drive the process and the sensitivity of target genes to these concentrations. In this paper we show how these quantities for a given transcription factor can be inferred from gene expression levels of a set of known target genes. We treat the protein concentration as a latent function with a Gaussian process prior, and include the sensitivities, mRNA decay rates and baseline expression levels as hyperparameters. We apply this procedure to a human leukemia dataset, focusing on the tumour repressor p53 and obtaining results in good accordance with recent biological studies.
4 0.69476891 191 nips-2006-The Robustness-Performance Tradeoff in Markov Decision Processes
Author: Huan Xu, Shie Mannor
Abstract: Computation of a satisfactory control policy for a Markov decision process when the parameters of the model are not exactly known is a problem encountered in many practical applications. The traditional robust approach is based on a worstcase analysis and may lead to an overly conservative policy. In this paper we consider the tradeoff between nominal performance and the worst case performance over all possible models. Based on parametric linear programming, we propose a method that computes the whole set of Pareto efficient policies in the performancerobustness plane when only the reward parameters are subject to uncertainty. In the more general case when the transition probabilities are also subject to error, we show that the strategy with the “optimal” tradeoff might be non-Markovian and hence is in general not tractable. 1
5 0.60446304 59 nips-2006-Context dependent amplification of both rate and event-correlation in a VLSI network of spiking neurons
Author: Elisabetta Chicca, Giacomo Indiveri, Rodney J. Douglas
Abstract: Cooperative competitive networks are believed to play a central role in cortical processing and have been shown to exhibit a wide set of useful computational properties. We propose a VLSI implementation of a spiking cooperative competitive network and show how it can perform context dependent computation both in the mean firing rate domain and in spike timing correlation space. In the mean rate case the network amplifies the activity of neurons belonging to the selected stimulus and suppresses the activity of neurons receiving weaker stimuli. In the event correlation case, the recurrent network amplifies with a higher gain the correlation between neurons which receive highly correlated inputs while leaving the mean firing rate unaltered. We describe the network architecture and present experimental data demonstrating its context dependent computation capabilities. 1
6 0.55484825 187 nips-2006-Temporal Coding using the Response Properties of Spiking Neurons
7 0.46102002 99 nips-2006-Information Bottleneck Optimization and Independent Component Extraction with Spiking Neurons
8 0.4595409 162 nips-2006-Predicting spike times from subthreshold dynamics of a neuron
9 0.45752132 189 nips-2006-Temporal dynamics of information content carried by neurons in the primary visual cortex
10 0.45616052 18 nips-2006-A selective attention multi--chip system with dynamic synapses and spiking neurons
11 0.40103847 154 nips-2006-Optimal Change-Detection and Spiking Neurons
12 0.39165944 165 nips-2006-Real-time adaptive information-theoretic optimization of neurophysiology experiments
13 0.37244645 71 nips-2006-Effects of Stress and Genotype on Meta-parameter Dynamics in Reinforcement Learning
14 0.3628104 29 nips-2006-An Information Theoretic Framework for Eukaryotic Gradient Sensing
15 0.34538153 81 nips-2006-Game Theoretic Algorithms for Protein-DNA binding
16 0.33559775 16 nips-2006-A Theory of Retinal Population Coding
17 0.32028714 197 nips-2006-Uncertainty, phase and oscillatory hippocampal recall
18 0.30426136 49 nips-2006-Causal inference in sensorimotor integration
19 0.30350178 192 nips-2006-Theory and Dynamics of Perceptual Bistability
20 0.29952133 17 nips-2006-A recipe for optimizing a time-histogram