Nearest neighbors reveal fast and slow components of motor learning

Changes in behaviour resulting from environmental influences, development and learning 1 – 5 are commonly quantified on the basis of a few hand-picked features 2 – 4 , 6 , 7 (for example, the average pitch of acoustic vocalizations 3 ), assuming discrete classes of behaviours (such as distinct vocal syllables) 2 , 3 , 8 – 10 . However, such methods generalize poorly across different behaviours and model systems and may miss important components of change. Here we present a more-general account of behavioural change that is based on nearest-neighbour statistics 11 – 13 , and apply it to song development in a songbird, the zebra finch 3 . First, we introduce the concept of ‘repertoire dating’, whereby each rendition of a behaviour (for example, each vocalization) is assigned a repertoire time, reflecting when similar renditions were typical in the behavioural repertoire. Repertoire time isolates the components of vocal variability that are congruent with long-term changes due to vocal learning and development, and stratifies the behavioural repertoire into ‘regressions’, ‘anticipations’ and ‘typical renditions’. Second, we obtain a holistic, yet low-dimensional, description of vocal change in terms of a stratified ‘behavioural trajectory’, revealing numerous previously unrecognized components of behavioural change on fast and slow timescales, as well as distinct patterns of overnight consolidation 1 , 2 , 4 , 14 , 15  across the behavioral repertoire. We find that diurnal changes in regressions undergo only weak consolidation, whereas anticipations and typical renditions consolidate fully. Because of its generality, our nonparametric description of how behaviour evolves relative to itself—rather than to a potentially arbitrary, experimenter-defined goal 2 , 3 , 14 , 16 —appears well suited for comparing learning and change across behaviours and species 17 , 18 , as well as biological and artificial systems 5 . A new method for analysing change in high-dimensional data is based on nearest-neighbour statistics and is applied here to song dynamics during vocal learning in zebra finches, but could potentially be applied to other biological and artificial behaviours.

[1]  R. Kerr,et al.  Discovery of Brainwide Neural-Behavioral Maps via Multiscale Unsupervised Structure Learning , 2014, Science.

[2]  Michale S Fee,et al.  A basal ganglia-forebrain circuit in the songbird biases motor output to avoid vocal errors , 2009, Proceedings of the National Academy of Sciences.

[3]  Jerome H. Friedman,et al.  A New Graph-Based Two-Sample Test for Multivariate and Object Data , 2013, 1307.6294.

[4]  P. Mitra,et al.  How sleep affects the developmental learning of bird song , 2005, Nature.

[5]  J. Born,et al.  Sleep forms memory for finger skills , 2002, Proceedings of the National Academy of Sciences of the United States of America.

[6]  E. Bizzi,et al.  Consolidation in human motor memory , 1996, Nature.

[7]  Marc Peter Deisenroth,et al.  Deep Reinforcement Learning: A Brief Survey , 2017, IEEE Signal Processing Magazine.

[8]  Ryan P. Adams,et al.  Mapping Sub-Second Structure in Mouse Behavior , 2015, Neuron.

[9]  Laurens van der Maaten,et al.  Accelerating t-SNE using tree-based algorithms , 2014, J. Mach. Learn. Res..

[10]  Zahir Tari,et al.  A Survey of Clustering Algorithms for Big Data: Taxonomy and Empirical Analysis , 2014, IEEE Transactions on Emerging Topics in Computing.

[11]  Kristin Branson,et al.  Computational Analysis of Behavior. , 2016, Annual review of neuroscience.

[12]  P. Mitra,et al.  De novo establishment of wild-type song culture in the zebra finch , 2009, Nature.

[13]  D. Wolpert,et al.  Context-Dependent Decay of Motor Memories during Skill Acquisition , 2013, Current Biology.

[14]  C. Elemans,et al.  Vocal Motor Performance in Birdsong Requires Brain–Body Interaction , 2019, eNeuro.

[15]  H. Nusbaum,et al.  Consolidation during sleep of perceptual learning of spoken language , 2003, Nature.

[16]  J. Kruskal.  Multidimensional scaling by optimizing goodness of fit to a nonmetric hypothesis , 1964 .

[17]  R. Yuste,et al.  Comprehensive machine learning analysis of Hydra behavior reveals a stable basal behavioral repertoire , 2018, eLife.

[18]  C. E. Ho,et al.  A procedure for an automated measurement of song similarity , 2000, Animal Behaviour.

[19]  Ashesh K Dhawale,et al.  The Role of Variability in Motor Learning. , 2017, Annual review of neuroscience.

[20]  F. Nottebohm,et al.  Dynamics of the Vocal Imitation Process: How a Zebra Finch Learns Its Song , 2001, Science.

[21]  Michael S. Brainard,et al.  What songbirds teach us about learning , 2002, Nature.

[22]  David J. Anderson,et al.  Toward a Science of Computational Ethology , 2014, Neuron.

[23]  A. Karni,et al.  Daytime sleep condenses the time course of motor memory consolidation , 2007, Nature Neuroscience.

[24]  Gary F. Marcus,et al.  Songbirds work around computational complexity by learning song vocabulary independently of sequence , 2017, Nature Communications.

[25]  P. Slater,et al.  Bird Song: Biological Themes and Variations , 1995 .

[26]  Daniel Margoliash,et al.  Sleep and sensorimotor integration during early vocal learning in a songbird , 2008, Nature.

[27]  R. Stickgold,et al.  Dissociable stages of human memory consolidation and reconsolidation , 2003, Nature.

[28]  L. Paninski,et al.  The Spatiotemporal Organization of the Striatum Encodes Action Space , 2017, Neuron.

[29]  J. Krakauer,et al.  Consolidation of motor memory , 2006, Trends in Neurosciences.