Computational Models of Memory Search

Memory models

Serial learning

give participant a list of words to learn

Associative chaining theory: forward bias

1. learn associations between item and its neighbours

2. strength of association decay monotonically with increasing distance between item presentation

Positional encoding theory: no bias

1. learn representation of the item's position in the list (index)

2. position cues item: house -> position 1 -> position 2 -> shoe

a. associative and b. positional

Representational assumptions

Memory matrix

two-dimensional matrix, each column is a memory vector

static

Although one can model such memories as a vector function of time, theorists usually eschew this added complexity, adopting a unitization assumption that underlies nearly all modern memory models.

Localist models

each item vector has a single, unique, nonzero element

each element corresponds to a unique item in memory

Distributed models

features representing an item distributed across many or all of the elements

The unitization assumption dovetails nicely with the classic list recall method in which the presentation of known items constitutes the miniexperiences to be stored and retrieved. But one can also create sequences out of unitary items, and by recalling and reactivating these sequences of items, one can model memories that include temporal dynamics.

Multi-trace theory

This model assumes that each item vector (memory) occupies its own “address,” much like memory stored on a computer is indexed by an address in the computer’s random-access memory. Repeating an item does not strengthen its existing entry but rather lays down a new memory trace.

retrieval of encoded item create new memory trace

this model implies number of traces can increase without bound, but brain capacity is finite

if search for an item is serial would take forever, if parallel would cause high demand on nervous systems

Composite memories

composite storage model

for recognition memory

storage equation: $mt = αm_{t−1} + B_t f_t$

Rather than summing item vectors directly, which results in substantial loss of information, we can first expand an item’s representation into a matrix form, and then sum the resultant matrices.

Summed similarity

Pattern completion

Contextual coding

Associative models

Recognition and recall

Serial learning

Recall phenomena

Serial position effects

Contiguity and similarity effects

Recall errors

Inter-response times

Memory search models

Dual-store theory

Retrieved context theory

context and item are retrieval cue for each other

paper