Difference Between Rsa And Mvp Classification Analysis And The New Information

1392 Words Aug 11th, 2015 6 Pages
1.1 What is representational similarity analysis?

Representational similarity analysis (RSA) is an analysis framework builds on a rich

psychological and mathematical literature, in which multi-channel measures of neural activity

are quantitatively related to each other and to computational theory and behavior by comparing

RDMs. RDM is the representational dissimilarity matrix, which contains a cell for each pair of

experimental conditions. Each cell is a number reflecting the dissimilarity between the activity

patterns associated with the two conditions. The core of the of RSA is to use RDM as a signature

of the representations in brain regions and computational models (Kriegeskorte, Mur, &

Bandettini, 2008).

1.2 The
…show more content…
It can be involved in an experiment with a

large number of conditions, but does not like MVP classification analysis with requires

predefined stimulus conditions or categories, it can go without the predefined category labels.

3) RSA also has less theoretical hypothesis and is more data-driven. Because in MVP

classification analysis, it is typical to focus on a certain dimension (e.g., animate vs. inanimate)

from the experimental conditions to classify, but RSA does not need the dimension to do the

discrimination, and thus will be affected less by the theoretical bias.

1.2.2 New information:

In the paper of Haxby et al. (2014), they summarized the advantages of RSA in the

following two points: First, RSA can reveal that representations in different brain areas differ

even if MVP classification is equivalent in those areas; Second, by converting the locations of

response vectors from a set of feature coordinates to a set of distances between vectors, the

geometry of the representational space is now in a format that is not dependent on the feature

coordinate axes. Along with those two advantages, I will talk about the new information RSA

can bring to us with four aspects as stated in Kriegeskorte et al (2008):

1) Integration of computational modeling into the analysis of brain-activity data.

Because the information needed for the comparison
Open Document