polyarea.m function  (MathWorks Inc)

Journal: bioRxiv

Article Title: Selective control of working memory in prefrontal, parietal, and visual cortex

doi: 10.1101/2020.04.07.030718

Figure Lengend Snippet: Selection transforms task-relevant information into a common subspace. (A) Population response for selected colors (binned into 4 color bins, indicated by marker color) at different locations (upper vs. lower, indicated by marker shape). Population response is taken as the vector of mean firing rate of all recorded neurons before the cue (pre-cue, left; taken at 400 ms) and after the cue (post-cue, right; taken just prior to target onset, see methods for details). Responses are projected into a reduced dimensionality subspace defined by the first three principle components (PCs) of all 8 color/location pairs. Grey lines connect adjacent colors along the color wheel. Gray shaded region reflects the best fitting planes to each location (see methods for details). (B) Color representations for upper and lower items become correlated after selection. Line shows the mean correlation between the population representation for each color when it was presented/remembered in the ‘upper’ or ‘lower’ position, over time. Correlation was measured after subtracting the mean response at each location (see methods for details). Error bars reflect standard error of the mean. (C) Color planes (seen in A) become aligned after selection, reflected in an increase in the cosine of the angle between the two color planes around the time of cue onset. Black line shows the best-fitting logistic function. (D) Alignment of color representations before (left) and after (right) selection. Colored markers indicate vector of population firing rate for both upper and lower items (markers as in A). Here, all vectors are projected into the ‘lower’ subspace, defined by the first two PCs that maximally explain variance in the color of the lower item (defined in the full N-dimensional neural space on held-out data; see methods). Timepoints and markers are as in (A). (E) Timecourse of population responses to the color of the upper item, projected into the upper subspace defined before selection (left) and after selection (right). Upper subspaces were defined as in D, but for the upper item. (F) Before selection, color representations are better separated using the pre-cue subspace. After selection, colors are better separated in the post-cue subspace. Separability was measured as the area of the quadrilateral defined by the population vectors for each color, projected into either the pre-cue or post-cue subspaces (left and right columns in each plot; area averaged across upper and lower items). Subspaces are defined as in D and E. Violin plots show bootstrapped distributions. (G) Schematic of how selection transforms color representations. Initially, the colors of the upper and lower item are encoded in orthogonal subspaces specific to each item’s location. The selected item is then transformed into a common subspace, regardless of its initial location. (H) Upper and lower representations become aligned after selection (left column) but immediately after stimulus presentation during attention (right column). Histograms show bootstrapped distribution of the cosine of the angle between the best-fitting planes for the upper and lower stimuli in either an ‘early’ (150-350 ms post-stimulus offset) or ‘late’ (200-0 ms before color wheel onset) time period during the delay. Green lines indicate median values. · p < 0.10, * p < 0.05, ** p < 0.01, *** p < 0.001.

Article Snippet: To measure separability of the colors, we computed the area of this quadrilateral (polyarea.m function in MATLAB).

Techniques: Selection, Marker, Plasmid Preparation, Transformation Assay