MOSAIC Model for Sensorimotor Learning and Control

By Masahiko Haruno, D. M. Wolpert, Mitsuo Kawato. Neural Computation, 2001

765 citations recorded by the source corpus.

Subfields: Biomedical Engineering, Cognitive Neuroscience, Control and Systems Engineering

View Paper: https://www.semanticscholar.org/paper/8db9b60d037518dab5cff2ddae23cb38379f90fd

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Biomedical Engineering

In Cognitive Neuroscience

In Control and Systems Engineering

Citations in the corpus

Augmented visual, auditory, haptic, and multimodal feedback in motor learning: A review (2012, background): “…Gandolfo, & Mussa-Ivaldi, 1997; Oakley & O ’ Modhrain, 2005), internal versus external focus during task execution (e.g., Criscimagna-Hemminger, Donchin, Gazzaniga, & Shadmehr, 2003; Shadmehr & Moussavi, 2000), and internal movement representation (e.g., Haruno et al., 2001; Todorov, 2004).”
Large Scale Machine Learning (2004, methodology): “Haruno et al. (2001) also proposed an algorithm, in which each expert is associated to a generative model, again trained on the whole training set. In a divide-and-conquer approach introduced by Rida et al. (1999), the authors propose to first divide the training set using an unsupervised algorithm ”
Learning to generate articulated behavior through the bottom-up and the top-down interaction processes (2003, background): “One might predict that the experimental results would be either (1) a smooth phase transition, (2) getting stuck in some local minimum, or (3) diverse transient dynamics.”
Decoupling Dynamics and Reward for Transfer Learning (2018, methodology): “Agrawal et al. (2016); Haruno et al. (2001) also use an inverse model to regularize their forward model as an auxiliary task, but without decoupling, showing transfer in dynamics, and operates only in the state space.”
Model learning for robot control: a survey (2011, methodology): “Such multiple pairs of forward and inverse models can be learned directly from data using gradientdescent methods or expectation-maximization (Haruno et al. 2001).”
The role of prediction in social neuroscience (2012, methodology): “Frameworks for forward models of action have been formulated and operationalized in computational models, such as the MOSAIC model (Haruno et al., 2001), which rely on Bayesian inferential statistics.”
Physiologically Inspired Model for the Visual Recognition of Transitive Hand Actions (2013, neutral): “Relationship to other models Many other biologically-relevant computational models for goaldirected action recognition have focused on the role of motor representations (Haruno et al., 2001; Wolpert et al., 2003) and specifically of the mirror neuron system (Oztop and Arbib, 2002; Demiris and Simmon”
Scaling Large Learning Problems with Hard Parallel Mixtures (2002, background): “(4) Reconstruct N subsets: for each example (xt, yt), • sort the experts in descending order according to the values wi(xt), • assign the example to the first expert in the list which has less than (T/N+ 1) examples in order to ensure a balance between the experts.”
Computational principles of sensorimotor control that minimize uncertainty and variability (2007, background): “Thecerebellararchitectureappearstobehighlymodular (Oscarsson, 1979; Bower & Woolston, 1983), so it is possible that the cerebellum implements many forward and inverse models simultaneously (Wolpert & Kawato, 1998; Wolpert et al. 1998; Haruno et al. 2001).”
A Bayesian Rule for Adaptive Control based on Causal Interventions (2009, methodology): “The idea of breaking down a control problem into a superposition of controllers has been previously evoked in the context of “mixture of experts”-models like the MOSAIC-architecture Haruno et al. [2001].”
A Minimum Relative Entropy Principle for Learning and Acting (2008, methodology): “For the control case, the idea of using mixtures of expert-controllers has been previously evoked in models like the MOSAIC-architecture (Haruno et al., 2001).”
From motor to sensory processing in mirror neuron computational modelling (2010, methodology): “Likewise, in Haruno et al. (2001), Keysers and Perrett (2004), Ito and Tani (2004), Oztop et al. (2005), and Oztop and Arbib (2002), one presupposes that a view-independent description of the environment is computed and fed into mirror neuron mechanisms.”
Emergence of Functional Hierarchy in a Multiple Timescale Neural Network Model: A Humanoid Robot Experiment (2008, background): “There exist earlier studies on the computational modeling of functional hierarchy in sequences of motor primitives, representative examples being the ‘‘mixture of expert’’ model [16] and the ‘‘MOSAIC’’ model [17].”
Learning Visual Predictive Models of Physics for Playing Billiards (2015, background): “Models of physics and model based control Works such as (Jordan & Rumelhart, 1992; Wolpert et al., 1995; Haruno et al., 2001; Todorov & Ghahramani, 2003) proposed using internal models of the external world for planning actions.”
Human volition: towards a neuroscience of will (2008, background): “Teleological pull also appears in recent computational models that couple motor prediction to motor plannin”
How Working Memory and the Cerebellum Collaborate to Produce Creativity and Innovation (2007, methodology): “…of the resulting state-space of the dynamics model as a three-dimensional problem space of previously learned movements or conceptual thought patterns (like those, e.g., confabulated in repetitive working memory processing; see, e.g., Haruno et al., 1999, 2001; Imamizu et al., 2003; Ito, 1997).”