Cerebellar acceleration of learning in an evidence-accumulation task.

Code ↔ Paper

4 matches between paragraphs of the paper and lines of its authors' code, computed by the harvester (lexical-v1). Click a colored paragraph or line to see its counterpart.

The 4 matches · 1 of them tie a paragraph to a whole file, not to given lines: a weak match, whose lines are not tinted

[1] § STAR★METHODS › METHOD DETAILS › Drift-diffusion modeling ↔ src/model_likelihood.jl, lines 327–408 · score 0.72 · model likelihood, standard deviation, pulses, Hessian, lapse, bias
[2] § STAR★METHODS › METHOD DETAILS › Drift-diffusion modeling ↔ src/PBupsModel.jl, the whole file · a weak match · score 0.62 · PBupsModel, model likelihood, fitted, accumulate
[3] § STAR★METHODS › QUANTIFICATION AND STATISTICAL ANALYSIS › Effect size calculations ↔ src/model_likelihood.jl, lines 327–408 · score 0.59 · square root, Standard deviation
[4] § STAR★METHODS › METHOD DETAILS › Behavior experiments ↔ settings/param_handlers.py, lines 13–114 · score 0.51 · Anti biasing, ports, interval, 200 ms, reward, delay

Paper

Loaded from Europe PMC by your browser, not stored by OSCR: doi.org · Europe PMC

The paper is loaded when this pane is shown.

The paper is loaded by your browser, with JavaScript: read it at doi.org or on Europe PMC.

The authors' code

Julia · 409 lines · 14 KB · MIT · 2 matches

# Global variables
const epsilon = 10.0^(-10);
const dx = 0.25;
const dt = 0.02;
const total_rate = 40;
# === Upgrading from ForwardDiff v0.1 to v0.2
# instead of ForwardDiff.GradientNumber and ForwardDiff.HessianNumber,
# we will use ForwardDiff.Dual
convert(::Type{Float64}, x::ForwardDiff.Dual) = Float64(x.value)
function convert(::Array{Float64}, x::Array{ForwardDiff.Dual})
y = zeros(size(x));
for i in 1:prod(size(x))
y[i] = convert(Float64, x[i])
end
return y
end
immutable NumericPair{X,Y} <: Number
x::X
y::Y
end
Base.isless(a::NumericPair, b::NumericPair) = (a.x<b.x) || (a.x==b.x && a.y<b.y)
"""
function bin_centers = make_bins(B, dx, binN)
Makes a series of points that will indicate bin centers. The first and
last points will indicate sticky bins. No "bin edges" are made-- the edge
between two bins is always implicity at the halfway point between their
corresponding centers. The center bin is always at x=0; bin spacing
(except for last and first bins) is always dx; and the position
of the first and last bins is chosen so that |B| lies exactly at the
midpoint between 1st (sticky) and 2nd (first real) bins, as well as
exactly at the midpoint between last but one (last real) and last
(sticky) bins.
Playing nice with ForwardDiff means that the *number* of bins must be predetermined.
So this function will not actually set the number of bins; what it'll do is determine their
locations. To accomplish this separation, the function uses as a third parameter binN,
which should be equal to the number of bins with bin centers > 0, as follows:
binN = ceil(B/dx)
and then the total number of bins will be 2*binN+1, with the center one always corresponding
to position zero. Use non-differentiable types for B and dx for this to work.
"""
function make_bins{T}(bins::Vector{T}, B, dx::T, binN)
cnt = 1
for i=-binN:binN
bins[cnt] = i*dx
cnt = cnt+1
end
if binN*dx == B
bins[end] = B + dx
bins[1] = -B - dx
else
bins[end] = 2*B - (binN-1)*dx
bins[1] = -2*B + (binN-1)*dx
end
end
"""
function F = Fmatrix([sigma, lambda, c], bin_centers)
Uses globals
dt
dx
epsilon (=10.0^-10)
Returns a square Markov matrix of transition probabilities.
Plays nice with ForwardDiff-- that is why bin_centers is a global vector (so that the rem
operations that go into defining the bins, which ForwardDiff doesn't know how to deal with,
stay outside of this differentiable function)
sigma should be in (accumulator units) per (second^(1/2))
lambda should be in s^-1
c should be in accumulator units per second
bin_centers should be a vector of the centers of all the bins. Edges will be at midpoints
between the centers, and the first and last bin will be sticky.
dx is not used inside Fmatrix, because bin_centers specifies all we need to know.
dt *is* used inside Fmatrix, to convert sigma, lambda, and c into timestep units
"""
function Fmatrix{T}(F::AbstractArray{T,2},params::Vector, bin_centers)
sigma2 = params[1];
lam = params[2];
c = params[3];
sigma2_sbin = convert(Float64, sigma2)
# println(typeof(sigma2))
# println(size(sigma2))
# println(sigma2)
# println("Converted ", sigma2_sbin)
# println(typeof(sigma2_sbin))
n_sbins = max(70, ceil(10*sqrt(sigma2_sbin)/dx))
F[1,1] = 1;
F[end,end] = 1;
swidth = 5.*sqrt(sigma2_sbin)
sbinsize = swidth/n_sbins #sbins[2] - sbins[1]
base_sbins = collect(-swidth:sbinsize:swidth)
ps = exp(-base_sbins.^2/(2*sigma2)) # exp(Array) -> exp.(x)
ps = ps/sum(ps);
sbin_length = length(base_sbins)
binN = length(bin_centers)
mu = 0.
for j in 2:binN-1
if lam == 0
mu = bin_centers[j] + c*dt#(exp(lam*dt))
else
mu = (bin_centers[j] + c/lam)*exp(lam*dt) - c/lam
end
for k in 1:sbin_length
sbin = (k-1)*sbinsize + mu - swidth
if sbin <= bin_centers[1] #(bin_centers[1] + bin_centers[2])/2
F[1,j] = F[1,j] + ps[k]
elseif bin_centers[end] <= sbin#(bin_centers[end]+bin_centers[end-1])/2 <= sbins[k]
F[end,j] = F[end,j] + ps[k]
else # more condition
if (sbin > bin_centers[1] && sbin < bin_centers[2])
lp = 1; hp = 2;
elseif (sbin > bin_centers[end-1] && sbin < bin_centers[end])
lp = binN-1; hp = binN;
else
lp = floor(Int,((sbin-bin_centers[2])/dx)) + 2#find(bin_centers .<= sbins[k])[end]
hp = ceil(Int,((sbin-bin_centers[2])/dx)) + 2#lp+1#Int(ceil((sbins[k]-bin_centers[2])/dx) + 1);
end
if lp == hp
F[lp,j] = F[lp,j] + ps[k]
else
F[hp,j] = F[hp,j] + ps[k]*(sbin - bin_centers[lp])/(bin_centers[hp] - bin_centers[lp])
F[lp,j] = F[lp,j] + ps[k]*(bin_centers[hp] - sbin)/(bin_centers[hp] - bin_centers[lp])
end
end
end
end
end
"""
version with inter-click interval(ici) for c_eff_net / c_eff_tot (followed the matlab code)
(which was using dt for c_eff)
function logProbRight(params::Vector)
RightClickTimes vector with elements indicating times of right clicks
LeftClickTimes vector with elements indicating times of left clicks
Nsteps number of timesteps to simulate
Takes params
sigma_a = params[1]; sigma_s = params[2]; sigma_i = params[3];
lambda = params[4]; B = params[5]; bias = params[6];
phi = params[7]; tau_phi = params[8]; lapse = params[9]
Returns the log of the probability that the agent chose Right.
"""
function logProbRight(RightClickTimes::Array{Float64,1}, LeftClickTimes::Array{Float64,1}, Nsteps::Int
;sigma_a = 0.01, sigma_s_R = 0.01, sigma_s_L = 0.01,
sigma_i = 0.01, lambda = 0., B = 8., bias = 0.,
phi = 1., tau_phi = 0.02, lapse_R = 0.01, lapse_L = 0.01,
input_gain_weight = 0.5)
# now it considers 12p/9p function
# we can update the default parameter later
# check whether sigma_S_L/sigma_S_R are specified
# 1. set one of them as sigma_s
if xor(sigma_s_R!=0.01, sigma_s_L!=0.01)
if sigma_s_R!=0.01
sigma_s_L=sigma_s_R
elseif sigma_s_L!=0.01
sigma_s_R=sigma_s_L
end
end
# same for lapse
# check whether lapse_L/lapse_R are specified
# 1. set one of them as lapse
if xor(lapse_R!=0.01, lapse_L!=0.01)
if lapse_R!=0.01
lapse_L=lapse_R
elseif lapse_L!=0.01
lapse_R=lapse_L
end
end
# function logProbRight(params::Vector, RightClickTimes::Array{Float64,1}, LeftClickTimes::Array{Float64,1}, Nsteps::Int)
# sigma_a = params[1]; sigma_s_R = params[2]; sigma_s_L = params[3];
# sigma_i = params[4]; lambda = params[5]; B = params[6]; bias = params[7];
# phi = params[8]; tau_phi = params[9]; lapse_R = params[10]; lapse_L = params[11];
# input_gain_weight = params[12];
if isempty(RightClickTimes) RightClickTimes = zeros(0) end;
if isempty(LeftClickTimes ) LeftClickTimes = zeros(0) end;
NClicks = zeros(Int, Nsteps);
Lhere = zeros(Int, length(LeftClickTimes));
Rhere = zeros(Int, length(RightClickTimes));
for i in 1:length(LeftClickTimes)
Lhere[i] = ceil((LeftClickTimes[i]+epsilon)/dt)
end
for i in 1:length(RightClickTimes)
Rhere[i] = ceil((RightClickTimes[i]+epsilon)/dt)
end
for i in Lhere
NClicks[Int(i)] = NClicks[Int(i)] + 1
end
for i in Rhere
NClicks[Int(i)] = NClicks[Int(i)] + 1
end
# === Upgrading from ForwardDiff v0.1 to v0.2
# instead of using convert we can use floor(Int, ForwardDiff.Dual) and
# ceil(Int, ForwardDiff.Dual)
binN = ceil(Int, B/dx)#Int(ceil(my_B/dx))
binBias = floor(Int, bias/dx) + binN+1
binBias_hp = ceil(Int, bias/dx) + binN+1
if binBias<1 binBias = 1; end
if binBias>binN*2+1 binBias = binN*2+1; end
if binBias_hp<1 binBias_hp = 1; end
if binBias_hp>binN*2+1 binBias_hp = binN*2+1; end
bin_centers = zeros(typeof(dx), binN*2+1)
make_bins(bin_centers, B, dx, binN)
# Visualization
a_trace = zeros(length(bin_centers), Nsteps)
c_trace = zeros(1, Nsteps)
## biased lapse rate (lapse_R, lapse_L)
a0 = zeros(typeof(sigma_a),length(bin_centers))
a0[1] = lapse_L/2;
a0[end] = lapse_R/2;
a0[binN+1] = 1-lapse_L/2-lapse_R/2;
temp_l = [NumericPair(LeftClickTimes[i],-1) for i=1:length(LeftClickTimes)]
temp_r = [NumericPair(RightClickTimes[i],1) for i=1:length(RightClickTimes)]
allbups = sort!([temp_l; temp_r])
if phi == 1
c_eff = 1.
else
c_eff = 0.
end
cnt = 0
Fi = zeros(typeof(sigma_i),length(bin_centers),length(bin_centers))
Fmatrix(Fi,[sigma_i, 0., 0.0], bin_centers)
a = Fi*a0;
a_trace[:,1] = a;
F0 = zeros(typeof(sigma_a),length(bin_centers),length(bin_centers))
Fmatrix(F0,[sigma_a*dt, lambda, 0.0], bin_centers)
for i in 2:Nsteps
net_input = 0.
c_eff_tot = 0.
c_eff_net = 0.
net_i_input = 0.
net_c_input = 0.
if NClicks[i-1]==0
c_eff_tot = 0.
c_eff_net = 0.
a = F0*a
else
for j in 1:NClicks[i-1]
if cnt != 0 || j != 1
ici = allbups[cnt+j].x - allbups[cnt+j-1].x
c_eff = 1 + (c_eff*phi - 1)*exp(-ici/tau_phi)
c_eff_tot = c_eff_tot + c_eff
c_eff_net = c_eff_net + c_eff*allbups[cnt+j].y
## (input_gain_weight) 0 to 1 : 0-left, 1-right
net_c_input = (c_eff_tot+c_eff_net)/2 # right
net_i_input = c_eff_tot-net_c_input # left
net_input = 2*input_gain_weight*net_c_input - 2*(1-input_gain_weight)*net_i_input
elseif cnt==0 && j==1
ici = 0.
c_eff = 1 + (c_eff*phi - 1)*exp(-ici/tau_phi)
c_eff_tot = c_eff_tot + c_eff
c_eff_net = c_eff_net + c_eff*allbups[cnt+j].y
## (input_gain_weight) 0 to 1 : 0-left, 1-right
# 0.5 is neutral
net_c_input = (c_eff_tot+c_eff_net)/2 # right
net_i_input = c_eff_tot-net_c_input # left
net_input = 2*input_gain_weight*net_c_input - 2*(1-input_gain_weight)*net_i_input
end
if j == NClicks[i-1]
cnt = cnt+j
end
end
## biased params added
net_sigma = sigma_a*dt + (sigma_s_R*net_i_input)/total_rate + (sigma_s_L*net_c_input)/total_rate
F = zeros(typeof(net_sigma),length(bin_centers),length(bin_centers))
Fmatrix(F,[net_sigma, lambda, net_input/dt], bin_centers)
a = F*a
end
a_trace[:,i] = a
c_trace[i] = c_eff*exp(lambda*dt)
end
# likelihood of poking right
if binBias == binBias_hp
pright = sum(a[binBias+1:end])+a[binBias]/2
else
pright = sum(a[binBias+2:end]) +
a[binBias]*((bin_centers[binBias+1] - bias)/dx/2) +
a[binBias+1]*(0.5 + (bin_centers[binBias+1] - bias)/dx/2)
end
if pright-1 < epsilon && pright > 1
pright = 1
end
if pright < epsilon && pright > 0
pright = 0
end
return log(pright)
end
function LogLikelihood(RightClickTimes::Vector, LeftClickTimes::Vector, Nsteps::Int, rat_choice::Int
;kwargs...)
if rat_choice > 0
# println("Right")
return logProbRight(RightClickTimes, LeftClickTimes, Nsteps;
kwargs...)#make_dict(args, x)...)
elseif rat_choice < 0
# println("Left")
return log(1 - exp(logProbRight(RightClickTimes, LeftClickTimes, Nsteps;
kwargs...)))#make_dict(args, x)...)))
end
end
"""
function (LL, LLgrad, LLhessian, bin_centers, bin_times, a_trace) =
llikey(params, rat_choice, maxT=1, RightPulseTimes=[], LeftPulseTimes=[], dx=0.25, dt=0.02)
Computes the log likelihood according to Bing's model, and returns log likelihood, gradient, and hessian
params is a vector whose elements, in order, are
sigma_a square root of accumulator variance per unit time sqrt(click units^2 per second)
sigma_s standard deviation introduced with each click (will get scaled by click adaptation)
sigma_i square root of initial accumulator variance sqrt(click units^2)
lambda 1/accumulator time constant (sec^-1). Positive means unstable, neg means stable
B sticky bound height (click units)
bias where the decision boundary lies (click units)
phi click adaptation/facilitation multiplication parameter
tau_phi time constant for recovery from click adaptation (sec)
lapse 2*lapse fraction of trials are decided randomly
rat_choice should be either "R" or "L"
RETURNS:
"""
# function single_trial(params::Vector, RightClickTimes::Vector, LeftClickTimes::Vector, Nsteps::Int, rat_choice::Int, hess_mode=0::Int)
# function llikey(params::Vector)
# logLike(params, RightClickTimes, LeftClickTimes, Nsteps, rat_choice)
# end
# if hess_mode > 0
# result = HessianResult(params)
# ForwardDiff.hessian!(result, llikey, params);
# else
# result = GradientResult(params)
# ForwardDiff.gradient!(result, llikey, params);
# end
# LL = ForwardDiff.value(result)
# LLgrad = ForwardDiff.gradient(result)
# if hess_mode > 0
# LLhessian = ForwardDiff.hessian(result)
# end
# if hess_mode > 0
# return LL, LLgrad, LLhessian
# else
# return LL, LLgrad
# end
# end

model_likelihood.jl at commit 70fff04, under MIT · at the source

Overview

Authors: Marlies Oostland^1,2,3, Mikhail Kislin^1,4,3, Yuhang Chen¹, Tiffany Chen⁵, Sarah Jo Venditto¹, Ben Deverett⁶, Samuel S-H Wang^1,7

ORCID iDs: Marlies Oostland, Mikhail Kislin

Neuroscience Institute, Princeton University, Princeton, NJ, USA
Swammerdam Institute for Life Sciences, University of Amsterdam, Amsterdam, the Netherlands
These authors contributed equally
Present address: Albert Einstein College of Medicine, 1410 Pelham Parkway S, Kennedy 915, Bronx, NY, USA
Department of Neurological Surgery, University of California, San Francisco, San Francisco, CA, USA
Department of Anesthesiology, Stanford University Medical Center, Stanford, CA, USA
Lead contact

Institutions: Princeton University (United States); University of Amsterdam (Netherlands); University of California, San Francisco (United States); Stanford Medicine (United States)

Journal: Cell reports, volume 45, issue 4, article 117262

Dates: published online 16 April 2026; in print April 2026

Type: Research article · Language: English

License: CC BY

Identifiers: DOI 10.1016/j.celrep.2026.117262 · PMID 41999601 · PMCID PMC13198533 · OpenAlex W4200393806

Open access: gold, a free copy (OpenAlex)

Preprint: doi:10.1101/2021.12.23.474034

Status: code verified

Categories: mouse (organism), autism (population), cognitive (subfield)

Methods: Statistics, Machine learning, Preprocessing, fMRI & imaging, Single-unit activity, calcium imaging, Physiology & signal measures, Connectivity

Keywords: Cerebellum, Learning, Cognition, Anterior cingulate cortex, Autism, Behavioral State, Complex Spikes, Sensory Reactivity, Task Focus, Cp: Neuroscience

Topic: Neural dynamics and brain function (Cognitive Neuroscience, Neuroscience), according to OpenAlex

Funding: NINDS NIH HHS (R01 NS045193, U19 NS104648); National Institute of Neurological Disorders and Stroke (R01 NS045193, U19 NS104648); National Institute of Mental Health (R01 MH115750); NIMH NIH HHS (R01 MH115750); EU Framework Programme for Research and Innovation Marie Skłodowska-Curie Actions (844318); Horizon 2020 Framework Programme; National Institutes of Health

Citations: not cited yet (Europe PMC); 68 references in the paper

Research resources: RRID:AB_143165, Corticosterone ELISA Kit RRID:AB_2877626, RRID:AB_476866, Mouse: C57BL/6J RRID:IMSR_JAX:000664, RRID:IMSR_JAX:004146, Mouse: Tsc1flox/flox: Tsc1tm1Djk/J RRID:IMSR_JAX:005680, RRID:IMSR_JAX:010536, RRID:IMSR_JAX:012567, RRID:IMSR_JAX:030328, ScanImage 2015 RRID:SCR_014307, i-control™ microplate reader software RRID:SCR_024562, NDP.view2 Plus RRID:SCR_025177

Abstract

Cerebellar processing contributes to sensory salience, cognition, and behavioral flexibility. Here, we report that learning on a sensory evidence-accumulation task in mice is accelerated by cue-locked optogenetic stimulation of Purkinje cells but not by continuous optogenetic interference. Latent-state analysis revealed that accelerated learning was associated with enhanced focus on current over past trials. A cerebellum-specific transgenic autism model with disrupted Purkinje cell function also unexpectedly showed accelerated learning as well as enhanced reactivity to touch and auditory cues. Transgenic mice and wild-type mice receiving cue-locked stimulation showed prolonged sensory responses in Purkinje cell complex spikes and anterior cingulate cortex, and a subset of Purkinje cells in crus I showed on-task enhanced response to stimuli in wild-type mice. Sensory salience and task state may be different facets of a complex-spike-based mechanism for regulating brain-wide learning mechanisms. These findings potentially link cerebellum-dependent sensory salience with a global weak coherence account of autism.

Reproduced under the paper's license (CC BY), from the paper cited above.

Repositories

Its files are read in the Code ↔ Paper reader above, with 4 matches between paragraphs and lines of code.

wanglabprinceton/accumulating_puffs

License: Apache-2.0

State: the link answers, verified on 29 September 2026

Evidence: files inventoried

Commit: d3ed79c86cf663921412b608aa0287b4d34ee035, 13 June 2018

Languages: Python (43)

Size: 57 files, 43 scripts

Software Heritage: archived

Found in: the text, “Behavior experiments”

Holds: README, license file

Not found: CITATION.cff, environment file, tests, continuous integration, documentation

Tools: NumPy (22 files), pandas (9 files), Matplotlib (4 files), h5py (3 files), SciPy (1 file)

Availability: 1 check, the latest on 29 September 2026: the link answers

29 September 2026: the link answers

45 files

config_example.py, Python, 32 lines
expts/__init__.py, Python, 1 line
expts/controllers.py, Python, 547 lines
expts/routines.py, Python, 8 lines
expts/saver.py, Python, 175 lines
expts/session.py, Python, 506 lines
expts/trials.py, Python, 500 lines
expts/views.py, Python, 446 lines
hardware/__init__.py, Python, 10 lines
hardware/actuator.py, Python, 142 lines
hardware/analog_reader.py, Python, 169 lines
hardware/audio.py, Python, 39 lines
hardware/cameras.py, Python, 486 lines
hardware/communications.py, Python, 41 lines
hardware/daq.py, Python, 168 lines
hardware/daqold.py, Python, 187 lines
hardware/dummy.py, Python, 81 lines
hardware/light.py, Python, 49 lines
hardware/mp285.py, Python, 230 lines
hardware/ni845x.py, Python, 132 lines
hardware/opto.py, Python, 45 lines
hardware/valve.py, Python, 113 lines
run.py, Python, 9 lines
settings/__init__.py, Python, 2 lines
settings/conditions.py, Python, 20 lines
settings/constants.py, Python, 4 lines
settings/durations.py, Python, 6 lines
settings/levels.py, Python, 207 lines
settings/manipulations.py, Python, 24 lines
settings/param_handlers.py, Python, 114 lines, 1 match
settings/ratios.py, Python, 7 lines
settings/rules.py, Python, 12 lines
subjects/__init__.py, Python, 1 line
subjects/subjects.py, Python, 77 lines
util/__init__.py, Python, 4 lines
util/custom_time.py, Python, 7 lines
util/edit_hdfstore_file.py, Python, 29 lines
util/fix_corrupt_ar.py, Python, 106 lines
util/logs.py, Python, 27 lines
util/new_data_file.py, Python, 26 lines
util/notifications.py, Python, 45 lines
util/process_data.py, Python, 39 lines
util/tcpip.py, Python, 81 lines
LICENSE, License, 201 lines
README.md, Text, 7 lines

At the source: github.com/wanglabprinceton/accumulating_puffs

misun6312/PBupsModel.jl

License: MIT

State: the link answers, verified on 29 September 2026

Evidence: files inventoried

Commit: 70fff040d4134591f6729e5864bb6a92d260117e, 28 August 2018

Languages: Julia (8)

Size: 17 files, 8 scripts

Software Heritage: not archived

Found in: the text, “Drift-diffusion modeling”

Holds: README, license file, tests, continuous integration

Not found: CITATION.cff, environment file, documentation

Availability: 1 check, the latest on 29 September 2026: the link answers

29 September 2026: the link answers

10 files

src/PBupsModel.jl, Julia, 48 lines, 1 match
src/all_trials.jl, Julia, 124 lines
src/bbox_hessian_keyword_minimization.jl, Julia, 429 lines
src/constrained_parabolic_minimization.jl, Julia, 229 lines
src/data_handle.jl, Julia, 45 lines
src/model_likelihood.jl, Julia, 409 lines, 2 matches
src/model_optimization.jl, Julia, 228 lines
test/runtests.jl, Julia, 150 lines
LICENSE.md, License, 22 lines
README.md, Text, 154 lines

At the source: github.com/misun6312/PBupsModel.jl

The paper's code and data availability statement is in the Data section.

Tracing map

Proposed by the machine: these links were found in the paper and verified at the source, without human review. The map will receive a Zenodo DOI once one of the paper's authors has validated it with their ORCID.

What the map holds:

2 repositories of the authors' code, each at its verified commit, with its license and how the link was found in the paper;
51 scripts, each with its path and the digest of its content;
4 matches between paragraphs of the paper and lines of the code (method lexical-v1);
neither the text of the paper nor the code itself.

Its JSON (tracing-map.json) is deposited on Zenodo with its DOI once the map is validated.

Data

No dataset and no data link were found in the paper.

Data and code availability

All data that support the findings of this study will be made available upon reasonable request.

Code used for data acquisition is available at https://github.com/wanglabprinceton/accumulating_puffs.

Any additional information required to reanalyze the data reported in this paper is available from the lead contact upon request.

Reproduced under the paper's license (CC BY), from the paper cited above.

Versions

The history of this record: each version stored by the harvester or made by a correction of its authors or of the maintainers of its code, and what changed in its facts. The texts of the paper (its abstract, its availability statements) are not part of it; versions that changed only those are not listed.

Version 1, 29 September 2026: the first record

Recorded: type, language, journal, volume, issue, pages, dates, 7 authors, 10 keywords, 7 funders, 65 references, 12 RRIDs.

Cite

This paper

Oostland, M., Kislin, M., Chen, Y., Chen, T., Venditto, S. J., Deverett, B., & Wang, S. S.-H. (2026). Cerebellar acceleration of learning in an evidence-accumulation task. Cell reports, 45(4), 117262. https://doi.org/10.1016/j.celrep.2026.117262

BibTeX

@article{oostland2026cerebellar, author = {Oostland, Marlies and Kislin, Mikhail and Chen, Yuhang and Chen, Tiffany and Venditto, Sarah Jo and Deverett, Ben and Wang, Samuel S-H}, title = {{Cerebellar acceleration of learning in an evidence-accumulation task}}, journal = {Cell reports}, year = {2026}, month = apr, volume = {45}, number = {4}, pages = {117262}, publisher = {Cell Press}, issn = {2211-1247}, doi = {10.1016/j.celrep.2026.117262}, url = {https://doi.org/10.1016/j.celrep.2026.117262}, pmid = {41999601}, pmcid = {PMC13198533} }

RIS

TY - JOUR AU - Oostland, Marlies AU - Kislin, Mikhail AU - Chen, Yuhang AU - Chen, Tiffany AU - Venditto, Sarah Jo AU - Deverett, Ben AU - Wang, Samuel S-H TI - Cerebellar acceleration of learning in an evidence-accumulation task T2 - Cell reports J2 - Cell Rep PY - 2026 DA - 2026/04/16 VL - 45 IS - 4 SP - 117262 SN - 2211-1247 PB - Cell Press DO - 10.1016/j.celrep.2026.117262 UR - https://doi.org/10.1016/j.celrep.2026.117262 LA - en ER -

CSL-JSON

{ "id": "10.1016/j.celrep.2026.117262", "type": "article-journal", "title": "Cerebellar acceleration of learning in an evidence-accumulation task", "container-title": "Cell reports", "author": [ { "family": "Oostland", "given": "Marlies" }, { "family": "Kislin", "given": "Mikhail" }, { "family": "Chen", "given": "Yuhang" }, { "family": "Chen", "given": "Tiffany" }, { "family": "Venditto", "given": "Sarah Jo" }, { "family": "Deverett", "given": "Ben" }, { "family": "Wang", "given": "Samuel S-H" } ], "container-title-short": "Cell Rep", "volume": "45", "issue": "4", "page": "117262", "DOI": "10.1016/j.celrep.2026.117262", "PMID": "41999601", "PMCID": "PMC13198533", "ISSN": "2211-1247", "publisher": "Cell Press", "URL": "https://doi.org/10.1016/j.celrep.2026.117262", "language": "en", "issued": { "date-parts": [ [ 2026, 4, 16 ] ] } }

The tracing map gets a citation of its own once an author has validated it and it has a DOI.

Similar papers

The papers with a page that share the most with this one: the tools found in their code, their categories, datasets, cited references and authors, the rarest counting most.

[1] doi:10.1038/s41398-026-03952-4 [code]: Perineuronal nets in cerebellar nuclei neurons orchestrate social behaviour via regulation of neuronal activity in circuits innervated by the cerebellum.
Journal: Translational psychiatry
In common: SciPy, autism, mouse, 5 references
[2] doi:10.1073/pnas.2616911123 [code]: Cerebellar microcircuits enable robust evidence-based decisions through cortico-cerebellar coupling.
Journal: Proceedings of the National Academy of Sciences of the United States of America
In common: Matplotlib, NumPy, cognitive, 5 references
[3] doi:10.1371/journal.pcbi.1014585 [code]: Decoding behavior with minimal and interpretable agent models.
Journal: PLoS computational biology
In common: pandas, SciPy, Matplotlib, 1 other tool, cognitive, mouse, 3 references
[4] doi:10.1038/s42003-026-10034-0 [code]: Region- and cell type-specific changes in gene expression in the cerebellum after classical fear conditioning.
Journal: Communications biology
In common: pandas, SciPy, Matplotlib, 1 other tool, mouse, 3 references
[5] doi:10.1002/advs.202519479 [code]: Diminished Signal-to-Noise Ratio Disrupts Somatosensory Population Encoding and Drives Tactile Hyposensitivity in the Fmr1<sup>-/y</sup> Autism Model.
Journal: Advanced science (Weinheim, Baden-Wurttemberg, Germany)
In common: h5py, pandas, SciPy, 2 other tools, autism, cognitive, mouse, 1 reference
[6] doi:10.1038/s41467-026-72445-1 [code]: Perception and neural representation of intermittent odor stimuli in mice.
Journal: Nature communications
In common: h5py, pandas, SciPy, 2 other tools, mouse, 2 references
[7] doi:10.1038/s41467-026-72057-9 [code]: Sex-specific behavioral feedback modulates sensorimotor processing and drives flexible social behavior.
Journal: Nature communications
In common: h5py, pandas, SciPy, 2 other tools, 2 references
[8] doi:10.1126/sciadv.aed6417 [code]: Intrinsic timing, not temporal prediction, underlies ramping dynamics in visual and parietal cortex during passive behavior.
Journal: Science advances
In common: h5py, pandas, SciPy, 2 other tools, mouse, 2 references
[9] doi:10.1038/s41593-026-02314-z [code]: Low-dimensional population dynamics in the brainstem gate REM sleep.
Journal: Nature neuroscience
In common: h5py, pandas, SciPy, 2 other tools, mouse, 2 references
[10] doi:10.1038/s41467-026-75371-4 [code]: Cerebellar neural populations orchestrate dopamine reward signaling with single-trial precision.
Journal: Nature communications
In common: pandas, SciPy, Matplotlib, 1 other tool, mouse, 2 references

Contribute

The authors of this paper can claim it, correct its record and validate its tracing map, and the maintainers of its code (its owner, or a public member of its organization) correct what it says of their repository; anyone signed in can ask for its removal. Every request goes to OSCR's own machine, which answers it; your account page follows them.

Sign in with ORCID to claim this paper as one of its authors, correct its record or validate its tracing map: when the paper's metadata lists your ORCID iD, you are recognized at once. Maintainers of its code: sign in with GitHub, then claim the repository on your account page.

Request its removal

To ask OSCR to remove this record, the copies of its authors' scripts or its tracing map, use the removal request page: signed in, you say who you are, what to remove and why, then review and confirm the request. Published rules decide every request (how).

Discussion, reproductions, activity

Discussion: questions and error reports about this paper and its code, from signed-in readers and its authors. It opens with sign-in.

Reproductions: reports from readers who ran the authors' code: what they reproduced, with which environment, commit and data. It opens with sign-in.

Activity: what happens around this paper: new versions of its record, its map's validation, discussions and reproductions. It opens with sign-in.