MINERVA(;L=.8, s=.9, N=10, S=20, h=.8, c=0)

Model object for the MINERVA model model.

Parameters

• L: learning rate or encoding accuracy probability
• memory: a matrix containing memory traces where each row corresponds to a trace
• n_features: the number of features per trace
• n_traces: the number of traces in model
• c: decision criteria

References

Hintzman, D. L. (1988). Judgments of frequency and recognition model in a multiple-trace model model. Psychological Review, 95(4), 528.

REM(;memory, u=.9, c=.6, g=.4, n_steps=1)

A REM model object containing a memory array and parameters.

Arguments

• memory::AbstractArray: An array of item representations.
• u::Float64: Probability (between 0 and 1) that a given feature of an item is copied into LTM.
• c::Float64: Probability (between 0 and 1) that a given feature of an item is copied correctly into LTM. 1 - c probability of a feature being converted to a different integer sampled from geometric distribution.
• g::Float64: Probability of sample integers from a geometric distribution.
• n_steps::Int64: The number of timesteps, i.e. number of times an item is studied.

References

Shiffrin, R. M., & Steyvers, M. (1997). A model for recognition memory: REM—Retrieving Effectively From Memory. Psychonomic Bulletin & Review, 4(2), 145-166.

• Authors *

@author taylor "dot" curley@gatech.edu (tmc2737) @date 06/17/2021 @coauthor itsdfish @date 06/18/2021

compute_activation(M, p)

Computes activation value for each trace in matrix M for a given probe p.

• M: a matrix in which rows represent memory traces and columns represent feature values
• p: a vector of feature values representing a memory probe

compute_activations(model::REM, probe)

Computes activations for each trace in memory given a memory probe. Activation estimates based on Equations 3 and 4 from original paper. The final product of the calculations is a probability ratio of the probe activating an old item vs. a new item. The example provided is from Figure 1 of the paper.

Arguments

• model::REM: an REM model object containing memory array and parameters
• probe::AbstractArray: A vector of numbers representing the memory probe.
• memory::AbstractArray: A vector of numbers representing an item in memory.

Output

• ::Float64: Float value representing the estimated odds ratio.

Example

julia> probe = [6,1,1,3];
julia> memory = [0 2; 1 2; 0 1; 3 0];
julia> model = REM(;memory, g=.40, c=.70)
julia> compute_activations(model, probe)
2-element Vector{Float64}:
 10.580277777777777
  0.18450000000000003

compute_intensity(model::MINERVA, probe)

Computes echo intensity using a given memory probe.

• model: MINERVA model object
• probe: a vector of feature values representing a memory probe

compute_odds(activations)

Given a set of activations, what are the odds that the probe is old vs. new? An OR > 1 –> old item. The odds are computed simply as a mean across activation values.

Arguments

• activations::AbstractArray: A vector of activations.

Output

• ::Float64: Float value representing the odds of old vs. new.

Example

julia> activations = [10.58,0.18];
julia> compute_odds(activations)
5.38

compute_prob(activations)

Given a set of activations, what is the probability that the probe is old vs. new? An prob > .5 –> old item. The probability is based on the odds, which is the mean across activations.

Arguments

• activations::AbstractArray: A vector of activations.

Output

• ::Float64: Float value representing the odds of old vs. new.

Example

julia> activations = [10.58,0.18];
julia> compute_odds(activations)
5.38

compute_activation(M, p)

Computes simularity between memory traces in matrix M and memory probe p using the dot product.

• M: a matrix in which rows represent memory traces and columns represent feature values
• p: a vector of feature values representing a memory probe

encode!(model, stimuli)

Encodes a matrix of stimuli with accuracy governed by parameter L

• model: MINERVA model object
• stimuli: a matrix in which rows represent stimuli and columns represent features

encode!(model::REM, stimuli)

Encodes a degraded copy of stimuli into memory.

Arguments

• memory::AbstractArray: A vector or matrix of item representations.
• u::Float64: Probability (between 0 and 1) that a given feature of an item is copied into LTM.
• c::Float64: Probability (between 0 and 1) that a given feature of an item is copied correctly into LTM. 1 - c probability of a feature being converted to a different integer sampled from geometric distribution.
• g::Float64: Probability of sample integers from a geometric distribution.
• n_steps::Int64: The number of timesteps, i.e. number of times an item is studied.

Output

• encode!::AbstractArray: an array of imperfect copy of items.

Example

julia> model = REM(;memory=fill(0, 4, 4), g=.40, c=.70)
julia> stimuli = generate_stimuli(.3, 4, 4)
julia> encode!(model, stimuli)
4×4 Matrix{Int64}:
 0  0  4   3
 1  1  0  15
 4  0  0   1
 0  0  0   0

generate_stimuli(g, w, n)

Generates a matrix containing individual vectors of feature values corresponding to different items. Rows correspond to feature values and columns correspond to memory traces or "images".

Arguments

• g::Float64: Controls the probability of sampling integers from a geometric distribution (between 0 and 1). A higher g value represents common words (with common features) and lower g values represent words with less common words (with uncommon features).
• w::Int64: The number of non-zero values in a given item vector.
• n::Int64: The number of items in a study/test list.

Output

• outMat::AbstractArray: An w x n matrix of separate item vectors.

Example

julia> generate_stimuli(0.4, 4, 4)
4×4 Array{Float64,2}:
 6.0  1.0  0.0  1.0
 0.0  6.0  2.0  1.0
 0.0  0.0  1.0  0.0
 0.0  0.0  3.0  0.0

random_stimulus(model::MINERVA, n)

Generates n random stimuli with a specified number of features in model object. Returns a n by n_features matrix.

random_stimulus(model::MINERVA)

Generates random stimulus with a specified number of features in the model object.