principal component analysis (pca) using built in functions utilising singular value decomposition with mean centring  (MathWorks Inc)

Journal: Frontiers in Neuroscience

Article Title: Post-ischemic reorganization of sensory responses in cerebral cortex

doi: 10.3389/fnins.2023.1151309

Figure Lengend Snippet: The top-3 independent components of the spiking response of each trial type. (A) shows each stimulation type and the corresponding independent components over the trial time. Positive coefficients are correlated with spiking activity while negative coefficients are anti-correlated with spiking activity. In the scatter plots below, each component is shown as an axis and each trial is plotted as a point within the three dimensions. Exemplar trials are highlighted and shown in insets with spike rate over time. (B) shows how the component weights (boxes) scale the component shapes to describe the features of the mean firing rate of an example channel. The corresponding blue and green arrows point to the deviations in mean firing rate while the purple arrow and line generally indicate the background firing rate that are captured by the respective component and its weight. (C) shows the reconstruction (shaded yellow) of the mean spike rate of an example channel (black line) using the descriptive weightings of the independent components.

Article Snippet: We first applied principal components analysis (PCA; MATLAB R2017a + ‘pca’ function with ‘Algorithm’ parameter set to ‘svd’) to qualitatively describe the different types of evoked responses for each condition, applying a singular value decomposition to the mean channel spike rates separately for each stimulus type; then, using the groupings for which the same basis subspace could accurately reconstruct the original observations, we seeded a reconstructed-independent components analysis algorithm (r-ICA; MATLAB R2017a + ‘rica’ function from the Statistics and Machine Learning Toolbox) using the top-3 combined-basis eigenvectors to recover a basis for the sets of components described above ( ).

Techniques: Activity Assay

Journal: Frontiers in Neuroscience

Article Title: Post-ischemic reorganization of sensory responses in cerebral cortex

doi: 10.3389/fnins.2023.1151309

Figure Lengend Snippet: Combined independent component analysis of the sensory response and its modulation. (A) shows the mean weights of the components sorted by stimulation type and area which are displayed in (B) . Positive values point to the presence of that component in the response while negative values indicate an inverse relationship; the error bars show the standard error of the mean. (C) displays the prediction of area and lesion volume for component 2 and 3 scores by the GLME model as compared to a linear fit. (D) highlights the changes in the component scores between Solenoid (yellow) and ICMS + Solenoid trials (purple) for each channel in an experimental block of an exemplar animal. (E) shows the reconstructed rates for each stimulation type by area. The mean component scores were used to weight each component and reconstruct the average response in spiking to stimulation.

Article Snippet: We first applied principal components analysis (PCA; MATLAB R2017a + ‘pca’ function with ‘Algorithm’ parameter set to ‘svd’) to qualitatively describe the different types of evoked responses for each condition, applying a singular value decomposition to the mean channel spike rates separately for each stimulus type; then, using the groupings for which the same basis subspace could accurately reconstruct the original observations, we seeded a reconstructed-independent components analysis algorithm (r-ICA; MATLAB R2017a + ‘rica’ function from the Statistics and Machine Learning Toolbox) using the top-3 combined-basis eigenvectors to recover a basis for the sets of components described above ( ).

Techniques: Blocking Assay