Browse
Submit Data
Databases
API
Help

Dataset Information

0 Views

0 Connections

0 Citations

0 Reanalyses

0 Downloads

Omics score: 0

Curriculum learning for human compositional generalization.

ABSTRACT: Generalization (or transfer) is the ability to repurpose knowledge in novel settings. It is often asserted that generalization is an important ingredient of human intelligence, but its extent, nature, and determinants have proved controversial. Here, we examine this ability with a paradigm that formalizes the transfer learning problem as one of recomposing existing functions to solve unseen problems. We find that people can generalize compositionally in ways that are elusive for standard neural networks and that human generalization benefits from training regimes in which items are axis aligned and temporally correlated. We describe a neural network model based around a Hebbian gating process that can capture how human generalization benefits from different training curricula. We additionally find that adult humans tend to learn composable functions asynchronously, exhibiting discontinuities in learning that resemble those seen in child development.

SUBMITTER: Dekker RB

PROVIDER: S-EPMC9564093 | biostudies-literature | 2022 Oct

REPOSITORIES: biostudies-literature

ACCESS DATA

Json Xml

Publications

Curriculum learning for human compositional generalization.

Dekker Ronald B RB Otto Fabian F Summerfield Christopher C

Proceedings of the National Academy of Sciences of the United States of America 20221003 41

Generalization (or transfer) is the ability to repurpose knowledge in novel settings. It is often asserted that generalization is an important ingredient of human intelligence, but its extent, nature, and determinants have proved controversial. Here, we examine this ability with a paradigm that formalizes the transfer learning problem as one of recomposing existing functions to solve unseen problems. We find that people can generalize compositionally in ways that are elusive for standard neural ...[more]

PMID: 36191191

Similar Datasets

Decomposing dynamical subprocesses for compositional generalization.

Project description:A striking feature of human cognition is an exceptional ability to rapidly adapt to novel situations. It is proposed this relies on abstracting and generalizing past experiences. While previous research has explored how humans detect and generalize single sequential processes, we have a limited understanding of how humans adapt to more naturalistic scenarios, for example, complex, multisubprocess environments. Here, we propose a candidate computational mechanism that posits compositional generalization of knowledge about subprocess dynamics. In two samples (N = 238 and N = 137), we combined a novel sequence learning task and computational modeling to ask whether humans extract and generalize subprocesses compositionally to solve new problems. In prior learning, participants experienced sequences of compound images formed from two graphs' product spaces (group 1: G1 and G2, group 2: G3 and G4). In transfer learning, both groups encountered compound images from the product of G1 and G3, composed entirely of new images. We show that subprocess knowledge transferred between task phases, such that in a new task environment each group had enhanced accuracy in predicting subprocess dynamics they had experienced during prior learning. Computational models utilizing predictive representations, based solely on the temporal contiguity of experienced task states, without an ability to transfer knowledge, failed to explain these data. Instead, behavior was consistent with a predictive representation model that maps task states between prior and transfer learning. These results help advance a mechanistic understanding of how humans discover and abstract subprocesses composing their experiences and compositionally reuse prior knowledge as a scaffolding for new experiences.

| S-EPMC11573675 | biostudies-literature

Value generalization in human avoidance learning.

Project description:Generalization during aversive decision-making allows us to avoid a broad range of potential threats following experience with a limited set of exemplars. However, over-generalization, resulting in excessive and inappropriate avoidance, has been implicated in a variety of psychological disorders. Here, we use reinforcement learning modelling to dissect out different contributions to the generalization of instrumental avoidance in two groups of human volunteers (N = 26, N = 482). We found that generalization of avoidance could be parsed into perceptual and value-based processes, and further, that value-based generalization could be subdivided into that relating to aversive and neutral feedback - with corresponding circuits including primary sensory cortex, anterior insula, amygdala and ventromedial prefrontal cortex. Further, generalization from aversive, but not neutral, feedback was associated with self-reported anxiety and intrusive thoughts. These results reveal a set of distinct mechanisms that mediate generalization in avoidance learning, and show how specific individual differences within them can yield anxiety.

| S-EPMC5957527 | biostudies-literature

Natural language instructions induce compositional generalization in networks of neurons.

Project description:A fundamental human cognitive feat is to interpret linguistic instructions in order to perform novel tasks without explicit task experience. Yet, the neural computations that might be used to accomplish this remain poorly understood. We use advances in natural language processing to create a neural model of generalization based on linguistic instructions. Models are trained on a set of common psychophysical tasks, and receive instructions embedded by a pretrained language model. Our best models can perform a previously unseen task with an average performance of 83% correct based solely on linguistic instructions (that is, zero-shot learning). We found that language scaffolds sensorimotor representations such that activity for interrelated tasks shares a common geometry with the semantic representations of instructions, allowing language to cue the proper composition of practiced skills in unseen settings. We show how this model generates a linguistic description of a novel task it has identified using only motor feedback, which can subsequently guide a partner model to perform the task. Our models offer several experimentally testable predictions outlining how linguistic information must be represented to facilitate flexible and general cognition in the human brain.

| S-EPMC11537972 | biostudies-literature

Generalization of multisensory perceptual learning.

Project description:Life in a multisensory world requires the rapid and accurate integration of stimuli across the different senses. In this process, the temporal relationship between stimuli is critical in determining which stimuli share a common origin. Numerous studies have described a multisensory temporal binding window-the time window within which audiovisual stimuli are likely to be perceptually bound. In addition to characterizing this window's size, recent work has shown it to be malleable, with the capacity for substantial narrowing following perceptual training. However, the generalization of these effects to other measures of perception is not known. This question was examined by characterizing the ability of training on a simultaneity judgment task to influence perception of the temporally-dependent sound-induced flash illusion (SIFI). Results do not demonstrate a change in performance on the SIFI itself following training. However, data do show an improved ability to discriminate rapidly-presented two-flash control conditions following training. Effects were specific to training and scaled with the degree of temporal window narrowing exhibited. Results do not support generalization of multisensory perceptual learning to other multisensory tasks. However, results do show that training results in improvements in visual temporal acuity, suggesting a generalization effect of multisensory training on unisensory abilities.

| S-EPMC4802214 | biostudies-other

Human-like systematic generalization through a meta-learning neural network.

Project description:The power of human language and thought arises from systematic compositionality-the algebraic ability to understand and produce novel combinations from known components. Fodor and Pylyshyn1 famously argued that artificial neural networks lack this capacity and are therefore not viable models of the mind. Neural networks have advanced considerably in the years since, yet the systematicity challenge persists. Here we successfully address Fodor and Pylyshyn's challenge by providing evidence that neural networks can achieve human-like systematicity when optimized for their compositional skills. To do so, we introduce the meta-learning for compositionality (MLC) approach for guiding training through a dynamic stream of compositional tasks. To compare humans and machines, we conducted human behavioural experiments using an instruction learning paradigm. After considering seven different models, we found that, in contrast to perfectly systematic but rigid probabilistic symbolic models, and perfectly flexible but unsystematic neural networks, only MLC achieves both the systematicity and flexibility needed for human-like generalization. MLC also advances the compositional skills of machine learning systems in several systematic generalization benchmarks. Our results show how a standard neural network architecture, optimized for its compositional skills, can mimic human systematic generalization in a head-to-head comparison.

| S-EPMC10620072 | biostudies-literature

Systematic Human Learning and Generalization From a Brief Tutorial With Explanatory Feedback.

Project description:We investigate human adults' ability to learn an abstract reasoning task quickly and to generalize outside of the range of training examples. Using a task based on a solution strategy in Sudoku, we provide Sudoku-naive participants with a brief instructional tutorial with explanatory feedback using a narrow range of training examples. We find that most participants who master the task do so within 10 practice trials and generalize well to puzzles outside of the training range. We also find that most of those who master the task can describe a valid solution strategy, and such participants perform better on transfer puzzles than those whose strategy descriptions are vague or incomplete. Interestingly, fewer than half of our human participants were successful in acquiring a valid solution strategy, and this ability was associated with completion of high school algebra and geometry. We consider the implications of these findings for understanding human systematic reasoning, as well as the challenges these findings pose for building computational models that capture all aspects of our findings, and we point toward a role for learning from instructions and explanations to support rapid learning and generalization.

| S-EPMC10898786 | biostudies-literature

Anticipated variability increases generalization of predictive learning.

Project description:We show that learners generalized more broadly around the learned stimulus when they expected more variability between the learning set and the generalization set, as well as within the generalization set. Experiments 1 and 3 used a predictive learning task and demonstrated border perceptual generalization both when expected variability was manipulated explicitly via instructions (Experiment 1), and implicitly by increasing temporal distance to the anticipated application of learning (Experiment 3). Experiment 2 showed that expecting to apply learning in the more distant future increases expected variability in the generalization set. We explain the relation between expected variability and generalization as an accuracy-applicability trade-off: when learners anticipate more variable generalization targets, they "cast a wider net" during learning, by attributing the outcome to a broader range of stimuli. The use of more abstract, broader categories when anticipating a more distant future application aligns with Construal Level Theory of psychological distance.

| S-EPMC11380665 | biostudies-literature

Adaptive coupling influences generalization of sensorimotor learning.

Project description:Sensorimotor learning typically shows generalization from one context to another. Models of sensorimotor learning characterize this with a fixed generalization function that couples learning between contexts. Here we examine whether such coupling is indeed fixed or changes with experience. We examine the interaction between motor memories for novel dynamics during reciprocating, back and forth reaching movements. Subjects first experienced a force field for one movement direction and we used channel trials to assess generalization on the reciprocal movements. This showed minimal coupling such that errors experienced for one movement direction did not lead to adaptation for the other. However, after subjects had experienced a force field for both movement directions concurrently, a coupling developed between the corresponding motor memories. That is, on re-exposure for one direction there was a significant adaptation for movements in the other direction. The coupling was specific to the errors experienced, with minimal coupling when the errors had the opposite sign to those experienced during adaptation. We developed a state-space model in which the states for the two movement directions are represented by separate, yet potentially coupled learning processes. The coupling in the model controlled the extent to which each learning process was updated by the errors experienced on the other movement direction. We show that the coupling relies on a memory trace of the consecutive errors experienced for both movement directions. Our results suggest that the generalization of motor learning is an adaptive process, reflecting the relation between errors experienced across different movements.

| S-EPMC6264158 | biostudies-literature

Threat learning promotes generalization of episodic memory.

Project description:The ability to generalize from and distinguish between aversive memories and novel experiences is critical to survival. Previous research has revealed mechanisms underlying generalization of threat-conditioned defensive responses, but little is known about generalization of episodic memory for threatening events. Here we tested if aversive learning influences generalization of episodic memory for threatening events in human adults. Subjects underwent Pavlovian threat-conditioning in which objects from one category were paired with a shock and objects from a different category were unpaired. The next day, subjects underwent a recognition memory test that included old, highly similar, and entirely novel items from the shock-paired and shock-unpaired object categories. Results showed that items highly similar to those from the object category previously paired with shock were mistaken for old items more often than items from the shock-unpaired category. This finding indicates that threat learning promotes generalization of episodic memory, and is consistent with the idea that threat generalization is an active process that may be adaptive for avoiding a myriad of potential threats following an emotional experience. Enhanced generalization of aversive episodic memories may be maladaptive, however, when old threat memories are inappropriately reactivated in harmless situations, exemplified in a number of stress- and anxiety-related disorders. (PsycINFO Database Record (c) 2019 APA, all rights reserved).

| S-EPMC6642861 | biostudies-literature

Compositional clustering in task structure learning.

Project description:Humans are remarkably adept at generalizing knowledge between experiences in a way that can be difficult for computers. Often, this entails generalizing constituent pieces of experiences that do not fully overlap, but nonetheless share useful similarities with, previously acquired knowledge. However, it is often unclear how knowledge gained in one context should generalize to another. Previous computational models and data suggest that rather than learning about each individual context, humans build latent abstract structures and learn to link these structures to arbitrary contexts, facilitating generalization. In these models, task structures that are more popular across contexts are more likely to be revisited in new contexts. However, these models can only re-use policies as a whole and are unable to transfer knowledge about the transition structure of the environment even if only the goal has changed (or vice-versa). This contrasts with ecological settings, where some aspects of task structure, such as the transition function, will be shared between context separately from other aspects, such as the reward function. Here, we develop a novel non-parametric Bayesian agent that forms independent latent clusters for transition and reward functions, affording separable transfer of their constituent parts across contexts. We show that the relative performance of this agent compared to an agent that jointly clusters reward and transition functions depends environmental task statistics: the mutual information between transition and reward functions and the stochasticity of the observations. We formalize our analysis through an information theoretic account of the priors, and propose a meta learning agent that dynamically arbitrates between strategies across task domains to optimize a statistical tradeoff.

| S-EPMC5929577 | biostudies-literature

OmicsDI is part of the ELIXIR infrastructure

OmicsDI is an Elixir interoperability service. Learn more ›

Tweets

OmicsDI Databases

PRIDE
PeptideAtlas
MassIVE
JPOST Repository
Physiome Model Repository

EGA
EVA
ENA
LINCS
PAXDB
Cell Collective

MetaboLights
Metabolomics Workbench
MetabolomeExpress
GNPS
BioModels
FAIRDOMHub

ArrayExpress
dbGaP
ExpressionAtlas
GEO
NODE

Information

Databases
Help
API
Contact us
Code on GitHub
Terms of use
Submit Data