Feature sequential sampling models #139

Zooaal · 2025-02-19T15:25:52Z

New Features:

New Component "Drift_Component" which is used for simulating evidence accumulation process with different Models.
3 Models for the usage in "Drift_Component": LBA, DDM and Kelly Model.
Functions to simulate the models traces and response times.
New Onset "SequenceOnset" which can be used with SequenceDesign to define a Onset for each Component in a Sequence.
New Onset "DriftOnset" special onset for the Drift_Component.
simulate_interonset_distances Functions for the "DriftOnset" used to return the response times of the models used.
simulate_onsets Function for the "SequenceOnset" specifies how the onsets for the sequences are calculated defined by the Onset for each Component. Important here: Onsets are defined for the next Component. For Example "SCR"=> S->C, C->R and R->S
New File "sequentialSamplingModelSimulation.jl": Defines the Kelly Model and its simulation. Furthermore the simulation of the LBA and DDM are designed. In the Future more Models can be defined here.

Important Changes:

Added using of SepuentialSamplingModels Package

On Top I added unittests and docstrings for all new features

For better usability I have created two HowTo in the documentation:

One explains how to simulate an entire evidence accumulation process as a basis.
Another one explains how to use the KellyModel, how to define parameters in the design and how to simulate an overlap of components.

behinger

This is an intermediate review, I'm running out of time, but I didnt check everything yet

src/component.jl

src/onset.jl

src/component.jl

Wording DocString Co-authored-by: Benedikt Ehinger <[email protected]>

github-actions

Remaining comments which cannot be posted as a review comment to avoid GitHub Rate Limit

JuliaFormatter

[JuliaFormatter] _{reported by reviewdog 🐶}

UnfoldSim.jl/src/onset.jl

Line 370 in a0f9253

end

[JuliaFormatter] _{reported by reviewdog 🐶}

UnfoldSim.jl/src/sequentialSamplingModelSimulation.jl

Lines 29 to 38 in a0f9253

    
           drift_rate::Union{Real, String} # drift rate 
        
           event_onset::Union{Real, String} # onset(sensory evidence) 
        
           sensor_encoding_delay::Union{Real, String} # var(sensory encoding delay) 
        
           accumulative_level_noise::Union{Real, String} # accum level noise 
        
           boundary::Union{Real, String} # boundaryary height 
        
           motor_onset::Union{Real, String} # onset(motor) 
        
           motor_delay::Union{Real, String} # var(motor) 
        
           post_accumulation_duration_mean::Union{Real, String} # mean(post decision) 
        
           post_accumulation_duration_variability::Union{Real, String} # var(post decision) 
        
           CPPrampdownDur::Union{Real, String} # CPPrampdown duration

[JuliaFormatter] _{reported by reviewdog 🐶}

UnfoldSim.jl/src/sequentialSamplingModelSimulation.jl

Line 51 in a0f9253

CPPrampdownDur = 0.1

[JuliaFormatter] _{reported by reviewdog 🐶}

UnfoldSim.jl/src/sequentialSamplingModelSimulation.jl

Lines 53 to 55 in a0f9253

    
           return new(drift_rate, event_onset, sensor_encoding_delay, accumulative_level_noise, boundary, motor_onset, 
        
               motor_delay, post_accumulation_duration_mean, post_accumulation_duration_variability, 
        
               CPPrampdownDur)

[JuliaFormatter] _{reported by reviewdog 🐶}

UnfoldSim.jl/src/sequentialSamplingModelSimulation.jl

Lines 82 to 84 in a0f9253

    
           return Dict( 
        
               name => getfield(model, name) for name in fieldnames(typeof(model)) 
        
           )

[JuliaFormatter] _{reported by reviewdog 🐶}

UnfoldSim.jl/src/sequentialSamplingModelSimulation.jl

Lines 111 to 122 in a0f9253

    
           evidence = zeros(length(time_vec)); 
        
           evidence[time_vec .>= (model.event_onset+(rand(rng) -.5)*model.sensor_encoding_delay)] .= 1;  
        
           startAccT = time_vec[findfirst(evidence .== 1)]; 
        
           noise = vcat(zeros( 
        
               sum(time_vec .< startAccT)), 
        
               randn(rng,sum(time_vec .>=  startAccT)) .*  model.accumulative_level_noise .*sqrt(Δt)); 
        
           ev=evidence; 
        
           ev[time_vec .< startAccT] .= 0; # set to zero all of the evidence before accT 
        
           cum_evidence = cumsum(ev .* model.drift_rate .* Δt .+ noise,dims=1); # This is the cumulative differential evidence, just as in a 1d DDM.

[JuliaFormatter] _{reported by reviewdog 🐶}

UnfoldSim.jl/src/sequentialSamplingModelSimulation.jl

Lines 124 to 125 in a0f9253

    
           cum_evidence = abs.(cum_evidence); 
        
           dti = findfirst(cum_evidence .> model.boundary); # finding the sample point of threshold crossing of each, then will pick the earlier as the winner

[JuliaFormatter] _{reported by reviewdog 🐶}

UnfoldSim.jl/src/sequentialSamplingModelSimulation.jl

Lines 130 to 131 in a0f9253

	rt = time_vec[dti] + model.motor_onset + (rand(rng)-.5) * model.motor_delay;

[JuliaFormatter] _{reported by reviewdog 🐶}

UnfoldSim.jl/src/sequentialSamplingModelSimulation.jl

Line 133 in a0f9253

    
           post_acc_duration = model.post_accumulation_duration_mean .+ model.post_accumulation_duration_variability .* rand(rng);

[JuliaFormatter] _{reported by reviewdog 🐶}

UnfoldSim.jl/src/sequentialSamplingModelSimulation.jl

Line 137 in a0f9253

acc_stop_index = dti + (post_acc_duration÷Δt)|>Int;

[JuliaFormatter] _{reported by reviewdog 🐶}

UnfoldSim.jl/src/sequentialSamplingModelSimulation.jl

Lines 139 to 140 in a0f9253

    
           cum_evidence = abs.(cum_evidence); 
        
           if acc_stop_index<length(time_vec)

[JuliaFormatter] _{reported by reviewdog 🐶}

UnfoldSim.jl/src/sequentialSamplingModelSimulation.jl

Lines 142 to 143 in a0f9253

    
           tmp = cum_evidence[acc_stop_index] .- (1:(nT-acc_stop_index)) .*cum_evidence[acc_stop_index] .* (Δt ./ model.CPPrampdownDur) 
        
           cum_evidence[(acc_stop_index+1):end] .= max.(Ref(0), tmp);

[JuliaFormatter] _{reported by reviewdog 🐶}

UnfoldSim.jl/src/sequentialSamplingModelSimulation.jl

Line 145 in a0f9253

return rt/Δt, cum_evidence[1:end]

[JuliaFormatter] _{reported by reviewdog 🐶}

UnfoldSim.jl/src/sequentialSamplingModelSimulation.jl

Line 178 in a0f9253

    
           function trace_sequential_sampling_model(rng, component::Drift_Component, design::AbstractDesign)

[JuliaFormatter] _{reported by reviewdog 🐶}

UnfoldSim.jl/src/sequentialSamplingModelSimulation.jl

Line 184 in a0f9253

    
           model = component.model_type(; (key => parameters[key] for key in keys(parameters))...)

[JuliaFormatter] _{reported by reviewdog 🐶}

UnfoldSim.jl/src/sequentialSamplingModelSimulation.jl

Line 258 in a0f9253

rt = (time_steps[end] + model.τ)/Δt

[JuliaFormatter] _{reported by reviewdog 🐶}

UnfoldSim.jl/src/sequentialSamplingModelSimulation.jl

Line 263 in a0f9253

rt = (time_vec[end] + model.τ)/Δt

[JuliaFormatter] _{reported by reviewdog 🐶}

UnfoldSim.jl/src/sequentialSamplingModelSimulation.jl

Line 299 in a0f9253

rt = (time_steps[end] + model.τ)/Δt

[JuliaFormatter] _{reported by reviewdog 🐶}

UnfoldSim.jl/src/sequentialSamplingModelSimulation.jl

Line 306 in a0f9253

end

[JuliaFormatter] _{reported by reviewdog 🐶}

UnfoldSim.jl/test/component.jl

Lines 49 to 59 in a0f9253

    
           # Test UnfoldSim.simulate_component(rng, c::Drift_Component, design::AbstractDesign) 
        
           boundary = 1.0 
        
           model_parameter = UnfoldSim.create_kelly_parameters_dict(UnfoldSim.KellyModel(boundary=boundary)); 
        
           c = Drift_Component(simulate_component, 0:1/500:1.0, 1/500, KellyModel, model_parameter); 
        
           design_single = UnfoldSim.SingleSubjectDesign(conditions = Dict(:condition => [1])); 
        
           design_seq = UnfoldSim.SequenceDesign(design_single,"SCR_"); 
        
           result_traces = UnfoldSim.simulate_component(StableRNG(1),c,design_seq) 
        
           @test size(result_traces) == (501, 3) 
        
           @test any(result_traces .== 0) 
        
           @test any(result_traces .>= boundary)

[JuliaFormatter] _{reported by reviewdog 🐶}

UnfoldSim.jl/test/component.jl

Lines 61 to 67 in a0f9253

    
           # Test UnfoldSim.simulate_component(rng, c::Drift_Component, design::AbstractDesign) 
        
           boundary = 1.0 
        
           model_parameter = UnfoldSim.create_kelly_parameters_dict(UnfoldSim.KellyModel(boundary=boundary)); 
        
           c = Drift_Component(simulate_component, 0:1/500:1.0, 1/500, KellyModel, model_parameter); 
        
           design_single = UnfoldSim.SingleSubjectDesign(conditions = Dict(:drift_rate => [0.5, 0.8], :condition => [1])); 
        
           design_seq = UnfoldSim.SequenceDesign(design_single,"SCR_"); 
        
           result_traces = UnfoldSim.simulate_component(StableRNG(1),c,design_seq)

[JuliaFormatter] _{reported by reviewdog 🐶}

UnfoldSim.jl/test/component.jl

Lines 69 to 71 in a0f9253

    
           @test size(result_traces) == (501, 6) 
        
           @test any(result_traces .== 0) 
        
           @test any(result_traces .>= boundary)

[JuliaFormatter] _{reported by reviewdog 🐶}

UnfoldSim.jl/test/component.jl

Lines 73 to 81 in a0f9253

    
           # Test calculate_response_times_for_ssm(rng, component::Drift_Component, design::AbstractDesign) 
        
           model_parameter = UnfoldSim.create_kelly_parameters_dict(UnfoldSim.KellyModel()); 
        
           c = Drift_Component(simulate_component, 0:1/500:1.0, 1/500, KellyModel, model_parameter); 
        
           design_single = UnfoldSim.SingleSubjectDesign(conditions = Dict(:drift_rate => [0.5, 0.8], :condition => [1])); 
        
           design_seq = UnfoldSim.SequenceDesign(design_single,"SCR_"); 
        
           sub_design = UnfoldSim.SubselectDesign(design_seq, 'C') 
        
           result_rts = UnfoldSim.calculate_response_times_for_ssm(StableRNG(1),c,sub_design) 
        
           @test size(result_rts) == (2,) 
        
           @test isapprox(result_rts, [399.6903067274333, 388.89617910657597], atol=1e-8)

[JuliaFormatter] _{reported by reviewdog 🐶}

UnfoldSim.jl/test/component.jl

Lines 83 to 90 in a0f9253

    
           # Test get_model_parameter(rng, evt, d::Dict) 
        
           rng = StableRNG(1) 
        
           model_parameter = UnfoldSim.create_kelly_parameters_dict(UnfoldSim.KellyModel(drift_rate="drift_rate")); 
        
           c = Drift_Component(simulate_component, 0:1/500:1.0, 1/500, KellyModel, model_parameter); 
        
           drift_rates = [0.5, 0.8] 
        
           design_single = UnfoldSim.SingleSubjectDesign(conditions = Dict(:drift_rate => drift_rates, :condition => [1])); 
        
           events = UnfoldSim.generate_events(rng, design_single) 
        
           for (i, evt) in enumerate(eachrow(events))

[JuliaFormatter] _{reported by reviewdog 🐶}

UnfoldSim.jl/test/component.jl

Line 93 in a0f9253

end

[JuliaFormatter] _{reported by reviewdog 🐶}

UnfoldSim.jl/test/onset.jl

Line 109 in a0f9253

fs=500

[JuliaFormatter] _{reported by reviewdog 🐶}

UnfoldSim.jl/test/onset.jl

Lines 123 to 127 in a0f9253

    
               Dict('S'=>UniformOnset(width=0,offset=80), 
        
                   'C'=>DriftOnset(), 
        
                   'R'=>UniformOnset(width=0,offset=120))) 
        
           model_parameter = UnfoldSim.create_kelly_parameters_dict(UnfoldSim.KellyModel(drift_rate="drift_rate")); 
        
           c = UnfoldSim.Drift_Component(simulate_component, 0:1/500:1.0, 1/500, KellyModel, model_parameter);

[JuliaFormatter] _{reported by reviewdog 🐶}

UnfoldSim.jl/test/onset.jl

Lines 129 to 130 in a0f9253

    
           design_single = UnfoldSim.SingleSubjectDesign(conditions = Dict(:drift_rate => [0.5, 0.8], :condition => [1])); 
        
           design_seq = UnfoldSim.SequenceDesign(design_single,"SCR_");

[JuliaFormatter] _{reported by reviewdog 🐶}

UnfoldSim.jl/test/onset.jl

Line 132 in a0f9253

    
           simulation = UnfoldSim.Simulation(design_rep, components, sequence_onset, UnfoldSim.NoNoise())

[JuliaFormatter] _{reported by reviewdog 🐶}

UnfoldSim.jl/test/onset.jl

Lines 143 to 146 in a0f9253

    
               Dict('S'=>UniformOnset(width=0,offset=80), 
        
                   'C'=>(DriftOnset(), UniformOnset(width=0, offset=140)), 
        
                   'R'=>UniformOnset(width=0,offset=120))) 
        
           simulation = UnfoldSim.Simulation(design_rep, components, sequence_onset, UnfoldSim.NoNoise())

[JuliaFormatter] _{reported by reviewdog 🐶}

UnfoldSim.jl/test/sequentialSamplingModelSimulation.jl

Line 3 in a0f9253

Δt = 1/fs; # time step

[JuliaFormatter] _{reported by reviewdog 🐶}

UnfoldSim.jl/test/sequentialSamplingModelSimulation.jl

Line 5 in a0f9253

time_vec = 0:Δt:tEnd; # time base

[JuliaFormatter] _{reported by reviewdog 🐶}

UnfoldSim.jl/test/sequentialSamplingModelSimulation.jl

Lines 10 to 11 in a0f9253

	km = KellyModel(event_onset=assert_event_onset, drift_rate=assert_drift_rate);

[JuliaFormatter] _{reported by reviewdog 🐶}

UnfoldSim.jl/test/sequentialSamplingModelSimulation.jl

Line 18 in a0f9253

    
           result_rt, result_trace = UnfoldSim.KellyModel_simulate_cpp(rng, KellyModel(boundary=boundary), time_vec, Δt)

[JuliaFormatter] _{reported by reviewdog 🐶}

UnfoldSim.jl/test/sequentialSamplingModelSimulation.jl

Line 21 in a0f9253

@test isapprox(result_rt, 399.6903067274333, atol=1e-8)

[JuliaFormatter] _{reported by reviewdog 🐶}

UnfoldSim.jl/test/sequentialSamplingModelSimulation.jl

Line 25 in a0f9253

    
           result_sim_rt, result_sim_trace = UnfoldSim.SSM_Simulate(rng, KellyModel(), Δt, time_vec)

[JuliaFormatter] _{reported by reviewdog 🐶}

UnfoldSim.jl/test/sequentialSamplingModelSimulation.jl

Lines 32 to 35 in a0f9253

    
           model_parameter = UnfoldSim.create_kelly_parameters_dict(KellyModel(boundary=boundary)); 
        
           c = UnfoldSim.Drift_Component(simulate_component, time_vec, Δt, KellyModel, model_parameter); 
        
           design_single = UnfoldSim.SingleSubjectDesign(conditions = Dict(:drift_rate => [0.5, 0.8], :condition => [1])); 
        
           design_seq = UnfoldSim.SequenceDesign(design_single,"SCR_");

[JuliaFormatter] _{reported by reviewdog 🐶}

UnfoldSim.jl/test/sequentialSamplingModelSimulation.jl

Line 37 in a0f9253

    
           result_rts, result_traces = UnfoldSim.trace_sequential_sampling_model(rng, c, design_seq)

[JuliaFormatter] _{reported by reviewdog 🐶}

UnfoldSim.jl/test/sequentialSamplingModelSimulation.jl

Line 44 in a0f9253

    
           result_rt, result_trace = UnfoldSim.SSM_Simulate(deepcopy(rng), KellyModel(), Δt, time_vec)

[JuliaFormatter] _{reported by reviewdog 🐶}

UnfoldSim.jl/test/sequentialSamplingModelSimulation.jl

Line 47 in a0f9253

@test isapprox(result_rt, 399.6903067274333, atol=1e-8)

[JuliaFormatter] _{reported by reviewdog 🐶}

UnfoldSim.jl/test/sequentialSamplingModelSimulation.jl

Line 54 in a0f9253

@test isapprox(result_rt, 223.00000000000003, atol=1e-8)

[JuliaFormatter] _{reported by reviewdog 🐶}

UnfoldSim.jl/test/sequentialSamplingModelSimulation.jl

Line 60 in a0f9253

@test isapprox(result_rt, 397.0, atol=1e-8)

Co-authored-by: Benedikt Ehinger <[email protected]>

github-actions

Remaining comments which cannot be posted as a review comment to avoid GitHub Rate Limit

JuliaFormatter

[JuliaFormatter] _{reported by reviewdog 🐶}

UnfoldSim.jl/test/sequentialSamplingModelSimulation.jl

Line 3 in ff43378

Δt = 1/fs; # time step

[JuliaFormatter] _{reported by reviewdog 🐶}

UnfoldSim.jl/test/sequentialSamplingModelSimulation.jl

Line 5 in ff43378

time_vec = 0:Δt:tEnd; # time base

[JuliaFormatter] _{reported by reviewdog 🐶}

UnfoldSim.jl/test/sequentialSamplingModelSimulation.jl

Lines 10 to 11 in ff43378

	km = KellyModel(event_onset=assert_event_onset, drift_rate=assert_drift_rate);

[JuliaFormatter] _{reported by reviewdog 🐶}

UnfoldSim.jl/test/sequentialSamplingModelSimulation.jl

Line 18 in ff43378

    
           result_rt, result_trace = UnfoldSim.KellyModel_simulate_cpp(rng, KellyModel(boundary=boundary), time_vec, Δt)

[JuliaFormatter] _{reported by reviewdog 🐶}

UnfoldSim.jl/test/sequentialSamplingModelSimulation.jl

Line 21 in ff43378

@test isapprox(result_rt, 399.6903067274333, atol=1e-8)

[JuliaFormatter] _{reported by reviewdog 🐶}

UnfoldSim.jl/test/sequentialSamplingModelSimulation.jl

Line 25 in ff43378

    
           result_sim_rt, result_sim_trace = UnfoldSim.SSM_Simulate(rng, KellyModel(), Δt, time_vec)

[JuliaFormatter] _{reported by reviewdog 🐶}

UnfoldSim.jl/test/sequentialSamplingModelSimulation.jl

Lines 32 to 35 in ff43378

    
           model_parameter = UnfoldSim.create_kelly_parameters_dict(KellyModel(boundary=boundary)); 
        
           c = UnfoldSim.Drift_Component(simulate_component, time_vec, Δt, KellyModel, model_parameter); 
        
           design_single = UnfoldSim.SingleSubjectDesign(conditions = Dict(:drift_rate => [0.5, 0.8], :condition => [1])); 
        
           design_seq = UnfoldSim.SequenceDesign(design_single,"SCR_");

[JuliaFormatter] _{reported by reviewdog 🐶}

UnfoldSim.jl/test/sequentialSamplingModelSimulation.jl

Line 37 in ff43378

    
           result_rts, result_traces = UnfoldSim.trace_sequential_sampling_model(rng, c, design_seq)

[JuliaFormatter] _{reported by reviewdog 🐶}

UnfoldSim.jl/test/sequentialSamplingModelSimulation.jl

Line 44 in ff43378

    
           result_rt, result_trace = UnfoldSim.SSM_Simulate(deepcopy(rng), KellyModel(), Δt, time_vec)

[JuliaFormatter] _{reported by reviewdog 🐶}

UnfoldSim.jl/test/sequentialSamplingModelSimulation.jl

Line 47 in ff43378

@test isapprox(result_rt, 399.6903067274333, atol=1e-8)

[JuliaFormatter] _{reported by reviewdog 🐶}

UnfoldSim.jl/test/sequentialSamplingModelSimulation.jl

Line 54 in ff43378

@test isapprox(result_rt, 223.00000000000003, atol=1e-8)

[JuliaFormatter] _{reported by reviewdog 🐶}

UnfoldSim.jl/test/sequentialSamplingModelSimulation.jl

Line 60 in ff43378

@test isapprox(result_rt, 397.0, atol=1e-8)

Formatting Co-authored-by: github-actions[bot] <41898282+github-actions[bot]@users.noreply.github.com>

test/sequentialSamplingModelSimulation.jl

src/UnfoldSim.jl

src/component.jl

src/onset.jl

behinger

nicely done! The code is already in good shape.

I think either define our own UnfoldSequentialSamplingModel or use the SamplingModels.SSM1D maybe? I like the later approach even better, maybe other models can then easily be reused from SequentialSamplingModels.jl as well

src/onset.jl

src/sequentialSamplingModelSimulation.jl

Co-authored-by: Benedikt Ehinger <[email protected]>

test/sequentialSamplingModelSimulation.jl

behinger

very cool!! I will try to take this for a spin soon!

src/component.jl

src/sequentialSamplingModelSimulation.jl

src/onset.jl

src/sequentialSamplingModelSimulation.jl

Co-authored-by: Benedikt Ehinger <[email protected]>

github-actions · 2025-05-16T10:48:42Z

src/component.jl

 )
    @debug "3D Array"
    off = get_offset(component) - minoffset(simulation.components)
    @views epoch_data[:, 1+off:length(component)+off, :] .+=


[JuliaFormatter] _{reported by reviewdog 🐶}

Suggested change

@views epoch_data[:, 1+off:length(component)+off, :] .+=

@views epoch_data[:, (1+off):(length(component)+off), :] .+=

github-actions · 2025-05-16T10:48:42Z

src/component.jl

+ 360.1329203034039
+```
+"""
+function calculate_response_times_for_ssm(rng, component::DriftComponent, design::UnfoldSim.SubselectDesign)


[JuliaFormatter] _{reported by reviewdog 🐶}

Suggested change

function calculate_response_times_for_ssm(rng, component::DriftComponent, design::UnfoldSim.SubselectDesign)

function calculate_response_times_for_ssm(

rng,

component::DriftComponent,

design::UnfoldSim.SubselectDesign,

)

github-actions · 2025-05-16T10:48:42Z

src/onset.jl

+# Returns
+- `simulate_interonset_distances`: function call.
+"""
+UnfoldSim.simulate_interonset_distances(rng, onset::AbstractOnset, design::AbstractDesign, components::AbstractComponent) = UnfoldSim.simulate_interonset_distances(rng,onset,design)


[JuliaFormatter] _{reported by reviewdog 🐶}

Suggested change

UnfoldSim.simulate_interonset_distances(rng, onset::AbstractOnset, design::AbstractDesign, components::AbstractComponent) = UnfoldSim.simulate_interonset_distances(rng,onset,design)

UnfoldSim.simulate_interonset_distances(

rng,

onset::AbstractOnset,

design::AbstractDesign,

components::AbstractComponent,

) = UnfoldSim.simulate_interonset_distances(rng, onset, design)

github-actions · 2025-05-16T10:48:42Z

src/onset.jl

+# Returns
+- `Vector{Int}`: the generated onsets for the drift components in the SequenceDesign.
+"""
+function UnfoldSim.simulate_interonset_distances(rng, onset::DriftOnset, design::AbstractDesign, components::AbstractComponent)


[JuliaFormatter] _{reported by reviewdog 🐶}

Suggested change

function UnfoldSim.simulate_interonset_distances(rng, onset::DriftOnset, design::AbstractDesign, components::AbstractComponent)

function UnfoldSim.simulate_interonset_distances(

rng,

onset::DriftOnset,

design::AbstractDesign,

components::AbstractComponent,

)

github-actions · 2025-05-16T10:48:43Z

src/onset.jl

+function UnfoldSim.simulate_interonset_distances(rng, onset::Tuple, design::AbstractDesign, components::AbstractComponent)
+    return Int.(round.(reduce(.+, simulate_interonset_distances.(deepcopy(rng), onset, Ref(design), Ref(components)))))


[JuliaFormatter] _{reported by reviewdog 🐶}

Suggested change

function UnfoldSim.simulate_interonset_distances(rng, onset::Tuple, design::AbstractDesign, components::AbstractComponent)

return Int.(round.(reduce(.+, simulate_interonset_distances.(deepcopy(rng), onset, Ref(design), Ref(components)))))

function UnfoldSim.simulate_interonset_distances(

rng,

onset::Tuple,

design::AbstractDesign,

components::AbstractComponent,

)

return Int.(

round.(

reduce(

.+,

simulate_interonset_distances.(

deepcopy(rng),

onset,

Ref(design),

Ref(components),

),

),

),

)

github-actions · 2025-05-16T10:48:43Z

src/onset.jl

+# Returns
+- `Vector{Float64}`: the generated onsets for the end of a sequence in the SequenceDesign.
+"""
+function UnfoldSim.simulate_interonset_distances(rng, onset::Char, design::AbstractDesign, components::AbstractComponent)


[JuliaFormatter] _{reported by reviewdog 🐶}

Suggested change

function UnfoldSim.simulate_interonset_distances(rng, onset::Char, design::AbstractDesign, components::AbstractComponent)

function UnfoldSim.simulate_interonset_distances(

rng,

onset::Char,

design::AbstractDesign,

components::AbstractComponent,

)

github-actions · 2025-05-16T10:48:43Z

src/onset.jl

+    onset_counter = Dict()
+    for k in keys(onset.onset)
+        sub_design = UnfoldSim.SubselectDesign(simulation.design, k)
+        onsets_for_k = simulate_interonset_distances(deepcopy(rng), onset.onset[k], sub_design, simulation.components[k][1])


[JuliaFormatter] _{reported by reviewdog 🐶}

Suggested change

onsets_for_k = simulate_interonset_distances(deepcopy(rng), onset.onset[k], sub_design, simulation.components[k][1])

onsets_for_k = simulate_interonset_distances(

deepcopy(rng),

onset.onset[k],

sub_design,

simulation.components[k][1],

)

github-actions · 2025-05-16T10:48:43Z

src/onset.jl

+        push!(final_onsets, onset_map[evt_k][onset_counter[evt_k]])
+        onset_counter[evt_k] += 1
+    end
+    final_onsets = vcat(final_onsets[end], final_onsets[1:end-1])


[JuliaFormatter] _{reported by reviewdog 🐶}

Suggested change

final_onsets = vcat(final_onsets[end], final_onsets[1:end-1])

final_onsets = vcat(final_onsets[end], final_onsets[1:(end-1)])

behinger · 2025-05-16T11:25:08Z

hey, I know you are official "out", but I'll come back to your offer at some point to still answer some questions ;-)

Do you know why LBA needs vector of driftrates? That makes it a bit ugly, because I need to do:

    SingleSubjectDesign(conditions = Dict(:drift_rate => [[3], [5]], :condition => [1]))
# instead of
    SingleSubjectDesign(conditions = Dict(:drift_rate => [3,5], :condition => [1]))

We definitely should try to catch that somehow

edit
Okay, reading the LBA help helps... we should highlight this somehow though. I'll think about it

Zooaal · 2025-05-16T11:57:06Z

Hey, no problem, you are always welcome to ask, but I may not answer so quickly.

github-actions · 2025-06-14T12:52:32Z

src/sequentialSamplingModels.jl

+"""
+function KellyModel_simulate_cpp(rng, model::KellyModel, time_vec, Δt)
+    evidence = zeros(length(time_vec))
+    evidence[time_vec.>=rand(rng, model.event_onset_distribution)] .= 1


[JuliaFormatter] _{reported by reviewdog 🐶}

Suggested change

evidence[time_vec.>=rand(rng, model.event_onset_distribution)] .= 1

evidence[time_vec .>= rand(rng, model.event_onset_distribution)] .= 1

github-actions · 2025-06-14T12:52:32Z

src/sequentialSamplingModels.jl

+"""
+function SSM_Simulate(rng, model::KellyModel, sfreq, max_length)
+    Δt = 1 / sfreq
+    time_vec = 0:Δt:max_length*Δt


[JuliaFormatter] _{reported by reviewdog 🐶}

Suggested change

time_vec = 0:Δt:max_length*Δt

time_vec = 0:Δt:(max_length*Δt)

github-actions · 2025-06-14T12:52:32Z

src/sequentialSamplingModels.jl

+    end
+    evidence = evidence[1:max_length]
+    return rt, evidence
+end


[JuliaFormatter] _{reported by reviewdog 🐶}

Suggested change

end

end

github-actions · 2025-06-14T13:01:36Z

docs/literate/tutorials/simulateDriftOverlap.jl

+    ),
+)
+# Deconvolution sucessfully removed the stimulus and response overlaps from our estimates, but it kept the drift-diffusion aspect nearly untouched. It is important to stress the "nearly", because depending on the `delay` parameters, this can potentially lead to problematic situation where it is unclear whether the drift activity should be assigned to S or R. In most realistic situations we found to be safe.
+# Further note that the drift-diffusion amplitude can be reduced because it is split up between `S` and `R` events.`


[JuliaFormatter] _{reported by reviewdog 🐶}

Suggested change

# Further note that the drift-diffusion amplitude can be reduced because it is split up between `S` and `R` events.`

# Further note that the drift-diffusion amplitude can be reduced because it is split up between `S` and `R` events.`

Zooaal added 2 commits February 13, 2025 13:15

add: Sequential Sampling Model, Evidence Accumulation Simulation

ffc7094

update: major fixes

0324d96

Zooaal requested review from behinger and jschepers February 19, 2025 15:25

behinger changed the base branch from main to exponential-noise February 21, 2025 11:05

behinger marked this pull request as ready for review February 21, 2025 11:05

behinger changed the base branch from exponential-noise to fix#124 February 21, 2025 11:06

behinger reviewed Feb 21, 2025

View reviewed changes

Update src/component.jl

a0f9253

Wording DocString Co-authored-by: Benedikt Ehinger <[email protected]>

github-actions bot reviewed Feb 25, 2025

View reviewed changes

Update src/component.jl

ff43378

Co-authored-by: Benedikt Ehinger <[email protected]>

github-actions bot reviewed Feb 25, 2025

View reviewed changes

Zooaal and others added 4 commits February 25, 2025 11:06

Apply suggestions from code review

115bd41

Formatting Co-authored-by: github-actions[bot] <41898282+github-actions[bot]@users.noreply.github.com>

Apply suggestions from code review

6d0e09e

Formatting Co-authored-by: github-actions[bot] <41898282+github-actions[bot]@users.noreply.github.com>

fixture from code review

cf1a757

Update DriftComponent to use max_length instead of an time vector

7bdc931

github-actions bot reviewed Feb 26, 2025

View reviewed changes

test/sequentialSamplingModelSimulation.jl Show resolved Hide resolved

Zooaal and others added 3 commits February 28, 2025 18:20

changed two KellyModel Parameter names

844262e

Update CITATION.cff

c524e9b

Change citation title to paper title

0da7b23

behinger changed the base branch from fix#124 to main March 25, 2025 08:59

behinger changed the base branch from main to fix#124 March 25, 2025 08:59

jschepers added 8 commits March 25, 2025 15:17

added joss badge to readme

e3e8e54

Edit citation section in README

46d3e28

Add BibTeX entry to readme

ae90422

Format citation section in README

af929ef

Format citation section in README

8de419c

Minor change

3974473

Fix and update CITATION.cff file

9a7b332

Add JOSS paper as preferred citation

1c35c8d

behinger reviewed Mar 26, 2025

View reviewed changes

src/onset.jl Outdated Show resolved Hide resolved

behinger reviewed Mar 27, 2025

View reviewed changes

Zooaal and others added 2 commits April 3, 2025 19:21

Apply basic suggestions from code review

f020998

Co-authored-by: Benedikt Ehinger <[email protected]>

Update from code review

e68c5dc

github-actions bot reviewed Apr 14, 2025

View reviewed changes

test/sequentialSamplingModelSimulation.jl Show resolved Hide resolved

test/sequentialSamplingModelSimulation.jl Show resolved Hide resolved

test/sequentialSamplingModelSimulation.jl Show resolved Hide resolved

test/sequentialSamplingModelSimulation.jl Show resolved Hide resolved

behinger reviewed Apr 16, 2025

View reviewed changes

Zooaal and others added 4 commits April 20, 2025 12:36

Apply suggestions from code review

82834d2

Co-authored-by: Benedikt Ehinger <[email protected]>

add onset transform to Int and comment on Kelly Model String Fields

9c0cdf7

v0.4.1 - compat DSP and SignalAnalysis

b0500b0

merge

2e23679

github-actions bot reviewed May 16, 2025

View reviewed changes

behinger added 2 commits May 16, 2025 13:25

Kelly Model is no longer exported, fix

dae4cc9

added SequentialSamplingModels to docs

0e2657a

jschepers and others added 9 commits June 11, 2025 11:41

Update CI.yml to run CI for PR on all branches not just main

d48031c

Update unfold family table & redirect doc links

3ccfa77

Update CI.yml

192f7b3

Update CI.yml

acec4ff

merged tutorials, added Truncated, added urgency signal

07cc43a

merge main

ba1ee0b

fix merge conflict remainder in tests

04460ce

added empty docstrings to abstracttypes

d1ff5e8

fix tests, added warning tau=0, export KellyModel,LBA,DDM

bfa3e80

github-actions bot reviewed Jun 14, 2025

View reviewed changes

oops forgot to save

4a042b8

github-actions bot reviewed Jun 14, 2025

View reviewed changes

Base automatically changed from fix#124 to v4.0 October 29, 2025 14:48

	drift_rate::Union{Real, String} # drift rate
	event_onset::Union{Real, String} # onset(sensory evidence)
	sensor_encoding_delay::Union{Real, String} # var(sensory encoding delay)
	accumulative_level_noise::Union{Real, String} # accum level noise
	boundary::Union{Real, String} # boundaryary height
	motor_onset::Union{Real, String} # onset(motor)
	motor_delay::Union{Real, String} # var(motor)
	post_accumulation_duration_mean::Union{Real, String} # mean(post decision)
	post_accumulation_duration_variability::Union{Real, String} # var(post decision)
	CPPrampdownDur::Union{Real, String} # CPPrampdown duration

	return new(drift_rate, event_onset, sensor_encoding_delay, accumulative_level_noise, boundary, motor_onset,
	motor_delay, post_accumulation_duration_mean, post_accumulation_duration_variability,
	CPPrampdownDur)

	return Dict(
	name => getfield(model, name) for name in fieldnames(typeof(model))
	)

	evidence = zeros(length(time_vec));
	evidence[time_vec .>= (model.event_onset+(rand(rng) -.5)*model.sensor_encoding_delay)] .= 1;
	startAccT = time_vec[findfirst(evidence .== 1)];

	noise = vcat(zeros(
	sum(time_vec .< startAccT)),
	randn(rng,sum(time_vec .>= startAccT)) .* model.accumulative_level_noise .*sqrt(Δt));
	ev=evidence;
	ev[time_vec .< startAccT] .= 0; # set to zero all of the evidence before accT
	cum_evidence = cumsum(ev .* model.drift_rate .* Δt .+ noise,dims=1); # This is the cumulative differential evidence, just as in a 1d DDM.

	cum_evidence = abs.(cum_evidence);
	dti = findfirst(cum_evidence .> model.boundary); # finding the sample point of threshold crossing of each, then will pick the earlier as the winner

	cum_evidence = abs.(cum_evidence);
	if acc_stop_index<length(time_vec)

	tmp = cum_evidence[acc_stop_index] .- (1:(nT-acc_stop_index)) .cum_evidence[acc_stop_index] . (Δt ./ model.CPPrampdownDur)
	cum_evidence[(acc_stop_index+1):end] .= max.(Ref(0), tmp);

	# Test UnfoldSim.simulate_component(rng, c::Drift_Component, design::AbstractDesign)
	boundary = 1.0
	model_parameter = UnfoldSim.create_kelly_parameters_dict(UnfoldSim.KellyModel(boundary=boundary));
	c = Drift_Component(simulate_component, 0:1/500:1.0, 1/500, KellyModel, model_parameter);
	design_single = UnfoldSim.SingleSubjectDesign(conditions = Dict(:condition => [1]));
	design_seq = UnfoldSim.SequenceDesign(design_single,"SCR_");
	result_traces = UnfoldSim.simulate_component(StableRNG(1),c,design_seq)

	@test size(result_traces) == (501, 3)
	@test any(result_traces .== 0)
	@test any(result_traces .>= boundary)

	@test size(result_traces) == (501, 6)
	@test any(result_traces .== 0)
	@test any(result_traces .>= boundary)

	# Test calculate_response_times_for_ssm(rng, component::Drift_Component, design::AbstractDesign)
	model_parameter = UnfoldSim.create_kelly_parameters_dict(UnfoldSim.KellyModel());
	c = Drift_Component(simulate_component, 0:1/500:1.0, 1/500, KellyModel, model_parameter);
	design_single = UnfoldSim.SingleSubjectDesign(conditions = Dict(:drift_rate => [0.5, 0.8], :condition => [1]));
	design_seq = UnfoldSim.SequenceDesign(design_single,"SCR_");
	sub_design = UnfoldSim.SubselectDesign(design_seq, 'C')
	result_rts = UnfoldSim.calculate_response_times_for_ssm(StableRNG(1),c,sub_design)
	@test size(result_rts) == (2,)
	@test isapprox(result_rts, [399.6903067274333, 388.89617910657597], atol=1e-8)

	# Test get_model_parameter(rng, evt, d::Dict)
	rng = StableRNG(1)
	model_parameter = UnfoldSim.create_kelly_parameters_dict(UnfoldSim.KellyModel(drift_rate="drift_rate"));
	c = Drift_Component(simulate_component, 0:1/500:1.0, 1/500, KellyModel, model_parameter);
	drift_rates = [0.5, 0.8]
	design_single = UnfoldSim.SingleSubjectDesign(conditions = Dict(:drift_rate => drift_rates, :condition => [1]));
	events = UnfoldSim.generate_events(rng, design_single)
	for (i, evt) in enumerate(eachrow(events))

	Dict('S'=>UniformOnset(width=0,offset=80),
	'C'=>DriftOnset(),
	'R'=>UniformOnset(width=0,offset=120)))
	model_parameter = UnfoldSim.create_kelly_parameters_dict(UnfoldSim.KellyModel(drift_rate="drift_rate"));
	c = UnfoldSim.Drift_Component(simulate_component, 0:1/500:1.0, 1/500, KellyModel, model_parameter);

	Dict('S'=>UniformOnset(width=0,offset=80),
	'C'=>(DriftOnset(), UniformOnset(width=0, offset=140)),
	'R'=>UniformOnset(width=0,offset=120)))
	simulation = UnfoldSim.Simulation(design_rep, components, sequence_onset, UnfoldSim.NoNoise())

	@views epoch_data[:, 1+off:length(component)+off, :] .+=
	@views epoch_data[:, (1+off):(length(component)+off), :] .+=

		function UnfoldSim.simulate_interonset_distances(rng, onset::Tuple, design::AbstractDesign, components::AbstractComponent)
		return Int.(round.(reduce(.+, simulate_interonset_distances.(deepcopy(rng), onset, Ref(design), Ref(components)))))

	final_onsets = vcat(final_onsets[end], final_onsets[1:end-1])
	final_onsets = vcat(final_onsets[end], final_onsets[1:(end-1)])

	evidence[time_vec.>=rand(rng, model.event_onset_distribution)] .= 1
	evidence[time_vec .>= rand(rng, model.event_onset_distribution)] .= 1

	time_vec = 0:Δt:max_length*Δt
	time_vec = 0:Δt:(max_length*Δt)

	# Further note that the drift-diffusion amplitude can be reduced because it is split up between `S` and `R` events.`
	# Further note that the drift-diffusion amplitude can be reduced because it is split up between `S` and `R` events.`

Feature sequential sampling models #139

Are you sure you want to change the base?

Feature sequential sampling models #139

Uh oh!

Conversation

Zooaal commented Feb 19, 2025

Uh oh!

behinger left a comment

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

github-actions bot left a comment

Choose a reason for hiding this comment

Uh oh!

github-actions bot left a comment

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

behinger left a comment

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

behinger left a comment

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

github-actions bot May 16, 2025

Choose a reason for hiding this comment

Uh oh!

github-actions bot May 16, 2025

Choose a reason for hiding this comment

Uh oh!

github-actions bot May 16, 2025

Choose a reason for hiding this comment

Uh oh!

github-actions bot May 16, 2025

Choose a reason for hiding this comment

Uh oh!

github-actions bot May 16, 2025

Choose a reason for hiding this comment

Uh oh!

github-actions bot May 16, 2025

Choose a reason for hiding this comment

Uh oh!

github-actions bot May 16, 2025

Choose a reason for hiding this comment

behinger commented May 16, 2025 •

edited

Loading