Method of generating an adaptive partial report and apparatus for implementing the same

A method of generating an adaptive partial report (adaptive PR) for an observer with an apparatus comprising a display, a user interface, and a processor, the method comprising

• the processor characterizing an iconic memory decay function for the observer, the iconic memory decay function having a plurality of parameters, the characterization including determining a prior for the plurality of parameters;
• the processor determining a first stimulus for a first trial based on the prior of the parameters, the determined first stimulus being expected to lead to an information gain for the iconic memory decay function;
• the display generating the first stimulus for viewing by the observer;
• the user interface receiving input for the first trial and in response to the first stimulus;
• the processor revising a posterior distribution of parameter values for the parameters based on the received input;
• the processor determining a new stimulus for a next trial based on the revised posterior distribution, the determined new stimulus being expected to lead to additional information gain for the iconic memory decay function;
• the display generating the new stimulus for viewing by the observer;
• the user interface receiving new input for the next trial and in response to the new stimulus;
• and
• the processor revising the posterior distribution of the parameter values for the parameters based on the new input, wherein the processor revises the posterior distribution of the parameter values for the parameters based on the new input by Bayesian inference,
• wherein the Bayesian inference follow Bayes rule
• p
• t
• ⁡
• (
• v
• ⇀
• ❘
• r
• t
• )
• =
• p
• ⁡
• (
• r
• t
• ❘
• v
• ⇀
• )
• ∑
• [
• p
• t
• ⁡
• (
• v
• ⇀
• )
• ×
• p
• ⁡
• (
• r
• t
• ❘
• v
• ⇀
• )
• ]
• ×
• p
• t
• ⁡
• (
• v
• ⇀
• )
• where {right arrow over (v)}=(a0, a1, τ) represents parameters of the sensory memory decay function, pt({right arrow over (v)}) is the prior probability density function of {right arrow over (v)}, pt(rt/{right arrow over (v)}) is the likelihood of observing a response (correct or incorrect) given {right arrow over (v)}, rt is observer's response in trial t, pt({right arrow over (v)}|rt is the posterior distribution of {right arrow over (v)} after the t-th trial.

An apparatus for generating an adaptive partial report, the apparatus comprising:

• a display;
• a user interface
• a processor;
• and
• a non-transitory medium comprising instructions that when executed by the processor cause the processor to
• characterize an iconic memory decay function for an observer, the iconic memory decay function having a plurality of parameters, the characterization including determining a prior of parameter values for the plurality of parameters,
• determine a first stimulus for a first trial based on the prior of the parameters, the determined first stimulus being expected to lead to an information gain for the iconic memory decay function,
• cause the display to generate the first stimulus for viewing by the observer,
• receive from the user interface input for the first trial and in response to the first stimulus,
• revise posterior distribution of parameter values for the parameters based on the received input,
• determine a new stimulus for a next trial based on the revised parameter values, the determined new stimulus being expected to lead to additional information gain for the estimated iconic memory decay function,
• cause the display to generate the new stimulus for viewing by the observer,
• receive from the user interface new input for the next trial and in response to the new stimulus, and
• revise the posterior distribution of parameter values for the parameters based on the new input,
• wherein the non-transitory medium further comprises instructions that when executed by the processor cause the processor to
• revise the posterior distribution of the parameter values for the parameters based on the received input by Bayesian inference, and
• revise the posterior distribution of the parameter values for the parameters based on the new input by Bayesian inference,
• wherein the Bayesian inference follows Bayes rule
• p
• t
• ⁡
• (
• v
• ⇀
• /
• r
• t
• )
• =
• p
• ⁡
• (
• r
• t
• /
• v
• ⇀
• )
• ∑
• [
• p
• t
• ⁡
• (
• v
• ⇀
• )
• ×
• p
• ⁡
• (
• r
• t
• /
• v
• ⇀
• )
• ]
• ×
• p
• t
• ⁡
• (
• v
• ⇀
• )
• where {right arrow over (v)}=(a0, a1, τ) represents parameters of the sensory memory decay function, pt({right arrow over (v)}) is the prior probability density function of {right arrow over (v)}, pt(rt|{right arrow over (v)}) is the likelihood of observing a response (correct or incorrect) given {right arrow over (v)}, rt is observer's response in trial t, pt({right arrow over (v)}|rt) is the posterior distribution of {right arrow over (v)} after the t-th trial.