PubMed Health⌕ Search

SEARCH · PubMed Health

Results for “Reinforcement 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.

At least 1,189 records · Page 66Linked to original sources
PubMed1996

D1/D5 dopamine receptor activation increases the magnitude of early long-term potentiation at CA1 hippocampal synapses.

The role of the mesolimbic dopaminergic system in the reinforcement of learning suggests that dopamine should be able to modulate activity-dependent synaptic plasticity. We have examined the effect of D1/D5 agonists on early long-term potentiation (LTP) (40 min) in the CA1 region of hippocampal slices. D1/D5 agonists (+)bromo-APB, 6-chloro-PB, and dihydrexidine increased the magnitude of LTP in a synapse-specific manner (by approximately 10, 15, and 20%, respectively). This D1/D5 effect was mimicked by a low dose (10 microM) of the adenylyl cyclase activator forskolin. The D1/D5 antagonist (+)SCH 23390 reduced early LTP. In catecholamine-depleted slices, LTP was smaller by approximately 20-25% and could not be decreased further by D1/D5 antagonist. Under these conditions, D1/D5 agonist 6-chloro-PB and forskolin produced a larger enhancement of LTP (20-25%), restoring it to the control level. At the same dose, dideoxyforskolin did not affect early LTP. The D1/D5 agonist effect was completely blocked by the D1/D5 antagonist (+)SCH 23390. These results indicate that dopamine produces a synapse-specific enhancement of early LTP through D1/D5 receptors and cAMP.

Animals↗
PubMed2006

The role of motivation in the performance of conditioned reflex switching of a maze skill in response to substitution of food reward quality in ants of the species Myrmica rubra.

The characteristics of learning in ants - active Myrmica rubra foragers - were studied in a maze at different levels of colony carbohydrate food need with reinforcement consisting of carbohydrate (sugar syrup) or protein (pupae of Lasius niger ants). Measures of the maze skill during learning reinforced with syrup were somewhat worse than those during learning reinforced with pupae, especially in terms of time-based measures. Ants were able to modify the acquired conditioned reflex reaction when the quality of reinforcement changed. At high levels of food need (" hungry" colony), substitution of syrup with larvae and vice versa was followed by transformation of the previously formed skill; this occurred at both experimental periods (training and testing). At low levels of motivation ("sated colony"), the optimized maze habit formed with protein reinforcement was replaced in the test with carbohydrate reinforcement by a "stochastic," unoptimized behavior with a dominance of investigative activity. These experiments demonstrated that conditioned reflex switching can occur in higher social insects - ants - in which the different forms depend on the level of colony need for food and, respectively, on the level of social food-related motivation of forager ants. The special importance of using the switch activating the corresponding motivational system before changing the reinforcement food quality is emphasized.

Animals↗
PubMed1979

Responsiveness of pre- and non-delinquents to incentive aspects of a reinforcing event.

A learning task, a six-position modification of a Fellows (1967) sequence, was used to examine the differences in responsiveness to symbolic and tangible incentives with 24 pre-delinquent and 25 non-delinquent 15-yr.-old boys. Under the low-incentive condition, only the symbolic incentive was available while under the high-incentive condition a symbolic and material reinforcer was employed. The high-incentive condition had no significant effect on the non-delinquents but did significantly affect the delinquents. Also, the high-incentive delinquent group performed similarly to both non-delinquent groups. Recommendations for research are given.

Adolescent↗
PubMed2003

Partial reinforcement across trials impairs escape performance but spares place learning in the water maze.

We studied the effects of partial reinforcement on escape performance and place learning in the water maze. Rats given 50% reinforcement across trials (i.e. the escape platform was present only on odd trials) were compared to controls given 100% reinforcement (platform present on all trials). Control groups either received 8 or 4 trials per day, which was equal to either the total number of trials (100%-8) or reinforced escapes (100%-4) of the 50% group. Analysis of escape performance (latency) revealed that the 50% group was impaired relative to the 100%-8 group, but not the 100%-4 group, during the first 5 days of acquisition. The 50% group was impaired relative to both control groups on days 6-10 of overtraining. However, analyses of within-trial behavior (target annulus preference and thigmotaxis) on nonreinforced trials suggest that the 50% group did learn the location of the hidden platform (place information), in addition to a wall-based thigmotactic response. By dividing the 60s nonreinforced trials into three 20-s time bins, we were able to detect a significant preference for the target annulus early in the trial (bin 1 of trial 40 and bins 1-2 of trial 80). Further, there was a significant increase in time spent in the periphery of the pool, near the wall, in the last time bin of trial 40. Because the platform was in the middle zone, this behavior competed with a place response. We conclude that across-trial partial reinforcement procedures may promote response competition and mask evidence of place learning in addition to weakening escape performance late in training.

Animals↗
PubMed1978

Effect of pimozide on the improvement in learning produced by self-stimulation and by water reinforcement.

When rats self-stimulate immediately after the training trial of an appetitive task their performance on a retention test is improved the next day. In the present study, this improvement was blocked by pretraining injections of pimozide, a dopaminergic blocker. In a second experiment, injections of pimozide retarded learning on the same task when the learning was reinforced by drinking water, but had no effect on learning which occurred in the absence of a reinforcer. The data made the hypotheses that the animal's behavior was a result of an action of primozide on sensory or motor mechanisms, or that the drug produced state-dependent effects, highly unlikely. We concluded that neural systems involving dopamine mediate an effect of reinforcing events on behavior.

Animals↗
PubMed2003

Involuntary, unreinforced (pure) spatial learning is impaired by fimbria-fornix but not by dorsal hippocampus lesions.

Pure spatial learning occurs when rats acquire information about an environment while exploring it in the absence of reinforcers. We previously reported that voluntary, unreinforced exploration of a radial maze retards subsequent reinforced conditioned cue preference (CCP) learning in the same maze. In the present experiment, we examined the effects of involuntary, unreinforced pre-exposure to a radial maze. During pre-exposure, rats were moved by an experimenter between the ends of two arms of a radial maze five times in 30 min. This form of pre-exposure retarded CCP learning, whereas rats that were not pre-exposed and rats that were pre-exposed to a maze in a different room displayed normal CCP learning. These findings suggest that some information specific to the maze environment was acquired during involuntary unreinforced pre-exposure to it. In experiment 2, the retardation of reinforced CCP learning by involuntary unreinforced pre-exposure was eliminated by fimbria-fornix lesions made before pre-exposure but was unaffected by fimbria-fornix lesions made after pre-exposure but before training. Large neurotoxic lesions of the dorsal hippocampus made before pre-exposure had no effect on the retardation of CCP learning, but the rats with these lesions were impaired on a standard test of reinforced spatial learning in a water maze. The lesion effects in experiment 2 are similar to those previously reported for voluntary exploration and suggest that pure spatial learning may occur during both voluntary exploration of and involuntary exposure to an environment in the absence of reinforcers. Pure spatial learning can apparently occur with exposure to two different locations within an environment, but the rats do not have to move between the locations voluntarily. An intact fimbria-fornix is required for acquisition but not expression of this form of learning. The hippocampus is not involved in this form of learning.

Animals↗
PubMed1963

Two parameters of a temporally defined schedule of negative reinforcement.

Temporal schedules of negative reinforcement were derived in a manner analogous to those for positive reinforcement, and two parametric studies were carried out with two of the three defining variables. With some limiting values of the defining variables, escape schedules result and data obtained at these values were similar to those obtained in escape studies. At other values of the defining variables avoidance behavior was generated.

Avoidance Learning↗
PubMed2005

Reinforcer devaluation in palatability-based learned flavor preferences.

Rats exposed to simultaneous compounds of 1 neutral flavor with dilute (2%) sucrose and a 2nd flavor with dilute (2%) maltodextrin subsequently consumed both flavors in preference to a 3rd flavor that was never paired with a palatable taste. Brief training exposure under ad lib food and water minimized the post-ingestive effects of nutrients, emphasizing the contribution of palatability to these preferences. Devaluation of sucrose or maltodextrin by pairing with illness (Experiment 1) or sensory-specific satiety (Experiment 2) selectively reduced the preference for the flavor previously paired with the devalued reinforcer. Such reinforcer-specific devaluation effects suggest that palatability-based learned flavor preferences are under-pinned by a Pavlovian process whereby the cue flavor is associated with the taste of the concurrently consumed palatable reinforcer.

Analysis of Variance↗
PubMed2003

Learning in spiking neural networks by reinforcement of stochastic synaptic transmission.

It is well-known that chemical synaptic transmission is an unreliable process, but the function of such unreliability remains unclear. Here I consider the hypothesis that the randomness of synaptic transmission is harnessed by the brain for learning, in analogy to the way that genetic mutation is utilized by Darwinian evolution. This is possible if synapses are "hedonistic," responding to a global reward signal by increasing their probabilities of vesicle release or failure, depending on which action immediately preceded reward. Hedonistic synapses learn by computing a stochastic approximation to the gradient of the average reward. They are compatible with synaptic dynamics such as short-term facilitation and depression and with the intricacies of dendritic integration and action potential generation. A network of hedonistic synapses can be trained to perform a desired computation by administering reward appropriately, as illustrated here through numerical simulations of integrate-and-fire model neurons.

Action Potentials↗
PubMed1987

Drives, affects, behavior--and learning: approaches to a psychobiology of emotion and to an integration of psychoanalytic and neurobiologic thought.

Drawing on concepts and data from psychoanalysis, neuroscience, ethology, and experimental psychology, this paper suggests that affects are the prime movers of human motivation and have, in addition, these properties: they are neurophysiologically generated, sensationlike signals originating not in classical sensory systems, but in limbic structures; activation of the neurobiologic affect generators reliably triggers motor stereotype--like facial expression, posture, and tone of voice--which in turn transmit the physical data underlying empathy and emotional communication; and appetitively sought pleasant emotions-sensations and aversive dysphoric feelings serve respectively as the ultimate, "brain-synthesized" positive and negative reinforcers of emotional learning. The latter ideas point to derivative hypotheses: that parental nonverbal behavior and communication evoke in children the affects that guide character development and choice of defense, and that these processes are consistent with an affect-reinforced associative learning model comprising both operant and classical conditioning paradigms. Psychoanalysis as therapy, then, may mechanistically rely for its clinical effectiveness on extinctionlike processes in which new experience in the analyst-patient relationship leads to the modulation or extinction of initially crippling signal affects associatively linked to various other affects, appetites, and behaviors. Finally, the core hypotheses mesh both with recent findings from the cellular study of affect and learning mechanisms and with associative learning-unlearning assumptions variably explicit in psychoanalytic writings since the time of Freud's (1895) Project.

Affect↗