19 papers · ranked by Valyu relevance
Leidi Zhao, Yiwen Zhao, Siddharth Patil, Dylan Davies + 3 more
'Lu Lu' 'Bo Ouyang'] Abstract— Advanced motor skills are essential for robots to physically coexist with humans. Much research on robot dynamics and control has achieved success on hyper robot motor capabilities, but mostly through heavily case-specific engineering. Meanwhile, in terms of robot acquiring skills in a…
Pei Xu, Xiumin Shang, Victor Zordan, Ioannis Karamouzas
Besides being able to perform a range of composite motions, humans typically learn such motions in an incremental manner. For example, if we know how to walk, we should be able to quickly learn how to hold our phone while walking. There is no need to relearn walking from scratch. Based on this intuition, we propose an…
Hiroshi Makino
The ability to compose new skills from a preacquired behavior repertoire is a hallmark of biological intelligence. Although artificial agents extract reusable skills from past experience and recombine them in a hierarchical manner, whether the brain similarly composes a novel behavior is largely unknown. In the present…
Liu, Zidong, Xu, Zhuoyan + 4 more
Composing basic skills from simple tasks to accomplish composite tasks is crucial for modern intelligent systems. We investigate the in-context composition ability of language models to perform composite tasks that combine basic skills demonstrated in in-context examples. This is more challenging than the standard…
Eric Schulz, Joshua B. Tenenbaum, David Duvenaud, Maarten Speekenbrink + 1 more
How do people recognize and learn about complex functional structure? Taking inspiration from other areas of cognitive science, we propose that this is achieved by harnessing compositionality: complex structure is decomposed into simpler building blocks. We formalize this idea within the framework of Bayesian…
Gauthier Boeshertz, Claudia Clopath
The brain builds predictive models to plan future actions. These models generalize remarkably well to new environments, but it is unclear how neural circuits acquire this flexibility. Here, we show that compositional representations emerge in recurrent neural networks (RNNs) trained solely to predict future sensory…
Yuma Osako, Aineias Arango, Toshitake Asabuki
Animals flexibly combine learned behaviors into novel actions without practicing their combinations, yet the computational mechanisms that enable independently acquired computations to be expressed in parallel remain unclear. Here we show that feedback geometry during learning determines whether recurrent dynamics can…
Jin Hwa Lee, Stefano Sarao Mannelli, Andrew Saxe
Learning Authors: ['Jin Hwa Lee' 'Stefano Sarao Mannelli' 'Andrew Saxe'] Diverse studies in systems neuroscience begin with extended periods of training known as 'shaping' procedures. These involve progressively studying component parts of more complex tasks, and can make the difference between learning a task quickly…
Felix Wiewel, Bin Yang
—Deep Neural Networks (DNNs) suffer from a rapid decrease in performance when trained on a sequence of tasks where only data of the most recent task is available. This phenomenon, known as catastrophic forgetting, prevents DNNs from accumulating knowledge over time. Overcoming catastrophic forgetting and enabling…
Jacob Russin, Ellie Pavlick, Michael J. Frank
Human learning embodies a striking duality: sometimes, we appear capable of following logical, compositional rules and benefit from structured curricula (e.g., in formal education), while other times, we rely on an incremental approach or trial-and-error, learning better from curricula that are randomly interleaved.…
Catherine Ulrich, Jesse L. M. Wilkins
Background Students’ ability to construct and coordinate units has been found to have far-reaching implications for their ability to develop sophisticated understandings of key middle-grade mathematical topics such as fractions, ratios, proportions, and algebra, topics that form the base of understanding for most…
Ciro Civile
This study investigated the role of perceptual learning in the composite face effect (CFE), which is characterized by reduced accuracy in recognizing the top half of a face when it is combined with the bottom half of another face, particularly when the composite is upright and aligned, compared to when the two halves…
Ravi D. Mill, Michael W. Cole
During cognitive task learning, neural representations must be rapidly constructed for novel task performance, then optimized for robust practiced task performance. How the geometry of neural representations changes to enable this transition from novel to practiced performance remains unknown. We hypothesized that…
Lennart Luettgau, Tore Erdmann, Sebastijan Veselic, Kimberly L. Stachenfeld + 3 more
'Kimberly L. Stachenfeld' 'Zeb Kurth-Nelson' 'Rani Moran' 'Raymond J. Dolan'] Title: Significance Humans possess a remarkable ability to adapt rapidly and flexibly to novel situations, a key aspect of cognition. While past studies detail how we learn from single processes, we have limited understanding of how we…
Ramsey Issa, Robert Sorenson, Taylor D. Sparks
The discovery of new dental materials is typically a slow process due to high-dimensionality of the formulation space as well as the multiple competing objectives which must be optimized for a given application. Here, we lay out a strategy using active learning and Bayesian optimization that has led to the discovery of…
Authors not listed
Machine olfaction—the artificial replication of the sense of smell—faces significant challenges due to the absence of large, standardized training datasets. Unlike vision, language, and audio models, which benefit from extensive corpora such as ImageNet, GLUE, and AudioSet, olfaction lacks scaled equivalents and…
Paulo F. Carvalho, Robert L. Goldstone
Inductive category learning takes place across time. As such, it is not surprising that the sequence in which information is studied has an impact in what is learned and how efficient learning is. In this paper we review research on different learning sequences and how this impacts learning. We analyze different…
Jeff Guo, Vendy Fialková, Juan Diego Arango, Christian Margreitter + 4 more
Reinforcement learning (RL) is a powerful paradigm that has gained popularity across multiple domains. However, applying RL may come at a cost of multiple interactions between the agent and the environment. This cost can be especially pronounced when the single feedback from the environment is slow or computationally…
Derek van Tilborg, Francesca Grisoni
Deep learning is accelerating drug discovery. However, current approaches are often affected by limitations in the available data, e.g., in terms of size or molecular diversity. Active deep learning has an untapped potential for low-data drug discovery, as it allows to improve a model iteratively during the screening…