neural decoding

discrimination between stimulus values is studied using receiver operated characteristics, likelihood ratio tests, and the [blank] lemma

Neyman-Pearson

the [blank] curve plots the true positive rate against the false positive rate and the area under the curve is a metric of overall performance

receiver operating characteristic (ROC)

[blank] are used in neural decoding to compare nested models, such as testing the improvement gained by adding a new feature to a decoder

Likelihood ratio tests (LRTs)

the Neyman-Pearson lemma creates a classifier that minimizes the probability of misclassification ([blank] error) while keeping the probability of a false alarm (Type I error) below a specified threshold (significance level)

Type II

static parameters can be decoded using the vector method, Bayesian, maximum a posteriori, and maximum likelihood inference, the Fisher information, and the [blank] lower bound

Cramer-Rao

the vector method involves using a [blank] vector to decode a given movement or action

population

the [blank] method uses use probabilistic models to predict external stimuli or movements from neural activity by combining prior knowledge with neural data

Bayesian

The [blank] method is a Bayesian technique used in neural decoding to estimate the most probable stimulus or variable that caused a specific pattern of neural activity

Maximum a Posteriori (MAP)

Unlike maximum likelihood estimation, which only considers the probability of observing the neural data given a stimulus, the MAP method incorporates a [blank] probability that accounts for background knowledge about how the stimulus is distributed in the real world

prior

Fisher information quantifies the amount of information a neural response carries about a stimulus and provides a theoretical [blank] bound on the accuracy of any decoding method

lower

Cramer-Rao lower bound (CRLB) is a statistical theorem that uses [blank] information to establish a theoretical minimum for the variance of any unbiased estimator

Fisher

Fisher information is a measure of the "information" or [blank] of the neural response to changes in a stimulus

sensitivity

Wiener filtering can be used to reconstruct an approximation of a [blank]-varying stimulus from the spike train it evokes

time

The Wiener filter in neural decoding is a linear method that estimates a neural signal from noisy neural data (like spike trains) by minimizing the [blank] and uses multiple linear regression

mean squared error