Special issue on information theoretic learning.
Explore the source record for details and available documents.
Algorithms↗
SEARCH · PubMed Health
Results for “Probability Learning”
Explore indexed PubMed citations for clinical trials, systematic reviews and public health research. Read source abstracts and follow each citation to its original PubMed record.
Quote a phrase for an exact phrase match. Source license links do not imply unrestricted reuse.
[Prediction of transcription and genomic sequences].
Technological developments have enhanced DNA sequencing at genomic scale. On the basis of the resulting sequences, computational biologists now attempt to localise the most important functional regions, starting with genes, but also importantly the regulatory motifs and conditions controlling their expression. In a recent paper published in Cell, M.A. Beer and S. Tavazoie report the results obtained by combining statistical classifications (clustering) of transcriptome data (DNA chips), software for the discovery of cis-regulatory patterns, together with a probabilistic learning method to infer regulatory rules tentatively accounting for the observed transcriptional profiles.
Forecasting↗
Improving examples to improve transfer to novel problems.
People often memorize a set of steps for solving problems when they study worked-out examples in domains such as math and physics without learning what domain-relevant subgoals or subtasks these steps achieve. As a result, they have trouble solving novel problems that contain the same structural elements but require different, lower-level steps. In three experiments, subjects who studied example solutions that emphasized a needed subgoal were more likely to solve novel problems that required a new approach for achieving this subgoal than were subjects who did not learn this subgoal. This result suggests that research aimed at determining the factors that influence subgoal learning may be valuable in improving transfer from examples to novel problems.
Adult↗
Probabilistic contingency learning with limbic or prefrontal damage.
A fundamental capacity of the human brain is to learn relations (contingencies) between environmental stimuli and the consequences of their occurrence. Some contingencies are probabilistic; that is, they predict an event in some situations but not in all. Animal studies suggest that damage to limbic structures or the prefrontal cortex may disturb probabilistic learning. The authors studied the learning of probabilistic contingencies in amnesic patients with limbic lesions, patients with prefrontal cortex damage, and healthy controls. Across 120 trials, participants learned contingent relations between spatial sequences and a button press. Amnesic patients had learning comparable to that of control subjects but failed to indicate what they had learned. Across the last 60 trials, amnesic patients and control subjects learned to avoid a noncontingent choice better than frontal patients. These results indicate that probabilistic learning does not depend on the brain structures supporting declarative memory.
Adult↗
Learning of "single trial" and "contextual" information processing in an odd-ball paradigm.
Explore the source record for details and available documents.
Attention↗
Linear-Nonlinear-Poisson models of primate choice dynamics.
The equilibrium phenomenon of matching behavior traditionally has been studied in stationary environments. Here we attempt to uncover the local mechanism of choice that gives rise to matching by studying behavior in a highly dynamic foraging environment. In our experiments, 2 rhesus monkeys (Macacca mulatta) foraged for juice rewards by making eye movements to one of two colored icons presented on a computer monitor, each rewarded on dynamic variable-interval schedules. Using a generalization of Wiener kernel analysis, we recover a compact mechanistic description of the impact of past reward on future choice in the form of a Linear-Nonlinear-Poisson model. We validate this model through rigorous predictive and generative testing. Compared to our earlier work with this same data set, this model proves to be a better description of choice behavior and is more tightly correlated with putative neural value signals. Refinements over previous models include hyperbolic (as opposed to exponential) temporal discounting of past rewards, and differential (as opposed to fractional) comparisons of option value. Through numerical simulation we find that within this class of strategies, the model parameters employed by animals are very close to those that maximize reward harvesting efficiency.
Animals↗
Spatial interference and response control in sequence learning: the role of explicit knowledge.
In several sequence learning studies it has been suggested that response control shifts from the stimuli to some internal representation (i.e., motor program) through the learning process. The main questions addressed in this paper are whether this control shift is related to explicit knowledge and whether the formation of these internal representations depends on the stimulus attributes. In one experiment we compared the learning of a response sequence triggered by either spatial location or location symbol (left-right) by using a serial response task (SRT). Symbols were presented at either a centered or random location. The results showed that in the symbolic conditions the shift of response control correlated with the emergence of explicit knowledge. Only participants with complete explicit knowledge seemed to learn the sequence structure beyond probabilistic information (response time "RT" did not depend on the frequency of the response). Moreover, these participants were able to overcome, when needed, spatial interference (RT was the same for both spatially corresponding and non-corresponding trials). However, when spatial location was relevant, RT was always faster, especially for more frequent responses. These results suggest that the relevant stimulus dimension (location or symbol) seems to engage different sequence learning mechanisms.
Adult↗
Alcohol and error processing.
A recent study indicates that alcohol consumption reduces the amplitude of the error-related negativity (ERN), a negative deflection in the electroencephalogram associated with error commission. Here, we explore possible mechanisms underlying this result in the context of two recent theories about the neural system that produces the ERN - one based on principles of reinforcement learning and the other based on response conflict monitoring.
Alcohol Drinking↗
Variational learning and bits-back coding: an information-theoretic view to Bayesian learning.
The bits-back coding first introduced by Wallace in 1990 and later by Hinton and van Camp in 1993 provides an interesting link between Bayesian learning and information-theoretic minimum-description-length (MDL) learning approaches. The bits-back coding allows interpreting the cost function used in the variational Bayesian method called ensemble learning as a code length in addition to the Bayesian view of misfit of the posterior approximation and a lower bound of model evidence. Combining these two viewpoints provides interesting insights to the learning process and the functions of different parts of the model. In this paper, the problem of variational Bayesian learning of hierarchical latent variable models is used to demonstrate the benefits of the two views. The code-length interpretation provides new views to many parts of the problem such as model comparison and pruning and helps explain many phenomena occurring in learning.
Algorithms↗
Neural networks in petroleum engineering: a case study.
A multilayer neural network has been used for deciding which oil reservoir layer has to be perforated. Many network architectures were tested until we found those with the best generalization capability. The network performs better than human experts and its achievements are higher than the historical average in the test area. As in other applications of neural networks, the learning capability improves with more hidden neurons but the generalization does not.
Algorithms↗
Variational learning for Gaussian mixture models.
This paper proposes a joint maximum likelihood and Bayesian methodology for estimating Gaussian mixture models. In Bayesian inference, the distributions of parameters are modeled, characterized by hyperparameters. In the case of Gaussian mixtures, the distributions of parameters are considered as Gaussian for the mean, Wishart for the covariance, and Dirichlet for the mixing probability. The learning task consists of estimating the hyperparameters characterizing these distributions. The integration in the parameter space is decoupled using an unsupervised variational methodology entitled variational expectation-maximization (VEM). This paper introduces a hyperparameter initialization procedure for the training algorithm. In the first stage, distributions of parameters resulting from successive runs of the expectation-maximization algorithm are formed. Afterward, maximum-likelihood estimators are applied to find appropriate initial values for the hyperparameters. The proposed initialization provides faster convergence, more accurate hyperparameter estimates, and better generalization for the VEM training algorithm. The proposed methodology is applied in blind signal detection and in color image segmentation.
Algorithms↗
A fixed interval schedule in which the interval is initiated by a response.
The fixed interval schedule described requires the animal to initiate every time interval by making a response on a bar other than the one on which it is reinforced. This response, R(A), demarcates the postreinforcement pause (S(R)-R(A) interval) from the fixed interval pause (R(A)-R(B) interval) so that these pauses may be measured separately. Twelve rats and three monkeys, working in two-bar Skinner boxes, were trained and stabilized on this schedule. The resulting performances, presented for individual animals, are analyzed in terms of (1) the relative frequencies with which the animal waits various lengths of time between consecutive responses, (2) the relative frequencies with which various rates of responding appear, (3) the change in response rate throughout the fixed interval, (4) the average length of the postreinforcement pause, (5) the relative frequencies with which the animal waits different lengths of time between the R(A) and the first R(B), and (6) the average inter-response time as a function of the rank order in the fixed interval of the inter-response time. The joint interpretation of the several measures taken leads to the following conclusions: 1. The probability of an R(B) increases throughout the fixed interval. 2. The increase is discontinuous at the first R(B), at which point the probability increases sharply. 3. The frequency distributions of R(A)-R(B) pauses exhibit three discrete types of behavior with no intermediate cases. 4. The (main) mode of R(A)-R(B) interval length usually occurs just below the fixed interval requirement.
Animals↗
Learning to commit or avoid the base-rate error.
When predicting an event, people neglect overall frequencies (base rates) of various possibilities. We have previously shown that this base-rate occurs not only with word problems, but also in a procedure with repeated trials: a sample cue followed by two choice options, one of which the subject must choose, with feedback regarding the correctness of the choice. Perhaps this base-rate error depends on people's histories of matching physically similar items. In support of this suggestions, we begin by showing that the base-rate error is eliminated with physically unrelated items. We then show that when the relation between items is again arbitrary, but is a relation that subjects already know, the error reappears. Finally we show that teaching subjects new arbitrary relations reintroduces the error in later testing. These experiments demonstrate a fundamental base-rate error dependent on learned relationships without interference from re-existing associations between cues and options, predictions may be made more optimally.
Concept Formation↗
Intentional control and implicit sequence learning.
Sequence knowledge acquired by repeated exposure to targets in a speeded localization task was studied in 3 experiments that sought to test A. Destrebecqz and A. Cleeremans's (2001, 2003) claim that, under certain circumstances, the expression of such sequence knowledge cannot be brought under intentional control. In Experiment 1 participants were trained on either a deterministic or a probabilistic sequence and then performed a free-generation test under either inclusion or exclusion instructions. Participants were found to be capable of both expressing (inclusion) and avoiding expressing (exclusion) sequence knowledge. These results were confirmed in Experiment 2 with a more exact replication of Destrebecqz and Cleeremans's methodology. In Experiment 3 participants performed a trial-by-trial generation test under both inclusion and exclusion conditions after a much longer period of training. All the findings are consistent with the proposal that information acquired during sequence learning is explicit in nature.
Adult↗
Alcohol-induced impairment of behavioral control: effects on the alteration and suppression of prepotent responses.
OBJECTIVE: Alcohol use in humans is associated with aggression and other socially inappropriate behaviors. These adverse effects have been attributed to an acute impairment of behavioral control, and research findings indicate that inhibitory aspects of behavioral control might be particularly vulnerable to the effects of alcohol. The present study tested the degree to which alcohol-induced impairment of behavioral control is due to a specific impairment of inhibitory mechanisms or due to a general information processing deficit. METHOD: Forty subjects (29 men) performed a cued reaction time task before and after receiving 0.65 g/kg alcohol or a placebo. Subjects performed the task under conditions that differed in the type of response needed to maintain behavioral control: response-suppression and response-alteration. RESULTS: Alcohol impairment was observed when behavioral controlwas dependent on response-suppression, but no impairment was observed when control relied on response-alteration. CONCLUSIONS: The findings point to the susceptibility of inhibitory processes by showing that alcohol can be particularly detrimental to behavioral control in situations where prepotent responses must be completely suppressed. Evidence for alcohol-induced impairment of inhibitory functions could provide important clues about basic behavioral mechanisms by which alcohol disrupts such higher order cognitive processes as working memory, learning and decision making.
Adult↗
Recognition of multiunit neural signals.
An essential step in studying nerve cell interaction during information processing is the extracellular microelectrode recording of the electrical activity of groups of adjacent cells. The recording usually contains the superposition of the spike trains produced by a number of neurons in the vicinity of the electrode. It is therefore necessary to correctly classify the signals generated by these different neurons. This paper considers this problem, and a new classification scheme is developed, which does not require human supervision. A learning stage is first applied on the beginning portion of the recording to estimate the typical spike shapes of the different neurons. As for the classification stage, a method is developed, which specifically considers the case when spikes overlap temporally. The method minimizes the probability of error, taking into account the statistical properties of the discharges of the neurons. The method is tested on a real recording as well as on synthetic data.
Action Potentials↗
Advances on BYY harmony learning: information theoretic perspective, generalized projection geometry, and independent factor autodetermination.
The nature of Bayesian Ying-Yang harmony learning is reexamined from an information theoretic perspective. Not only its ability for model selection and regularization is explained with new insights, but also discussions are made on its relations and differences from the studies of minimum description length (MDL), Bayesian approach, the bit-back based MDL, Akaike information criterion (AIC), maximum likelihood, information geometry, Helmholtz machines, and variational approximation. Moreover, a generalized projection geometry is introduced for further understanding such a new mechanism. Furthermore, new algorithms are also developed for implementing Gaussian factor analysis (FA) and non-Gaussian factor analysis (NFA) such that selecting appropriate factors is automatically made during parameter learning.
Algorithms↗
Combining exemplar-based category representations and connectionist learning rules.
Adaptive network and exemplar-similarity models were compared on their ability to predict category learning and transfer data. An exemplar-based network (Kruschke, 1990a, 1990b, 1992) that combines key aspects of both modeling approaches was also tested. The exemplar-based network incorporates an exemplar-based category representation in which exemplars become associated to categories through the same error-driven, interactive learning rules that are assumed in standard adaptive networks. Experiment 1, which partially replicated and extended the probabilistic classification learning paradigm of Gluck and Bower (1988a), demonstrated the importance of an error-driven learning rule. Experiment 2, which extended the classification learning paradigm of Medin and Schaffer (1978) that discriminated between exemplar and prototype models, demonstrated the importance of an exemplar-based category representation. Only the exemplar-based network accounted for all the major qualitative phenomena; it also achieved good quantitative predictions of the learning and transfer data in both experiments.
Concept Formation↗