Demixed principal component analysis of neural population data

By Dmitry Kobak, Wieland Brendel, Christos Constantinidis, C. Feierstein, Adam Kepecs, Zachary F. Mainen, Xue-Lian Qi, Ranulfo Romo, Naoshige Uchida, Christian K. Machens. eLife, 2014

567 citations recorded by the source corpus.

Neurons in higher cortical areas, such as the prefrontal cortex, are often tuned to a variety of sensory and motor variables, and are therefore said to display mixed selectivity. This complexity of single neuron responses can obscure what information these areas represent and how it is represented. Here we demonstrate the advantages of a new dimensionality reduction technique, demixed principal component analysis (dPCA), that decomposes population activity into a few components. In addition to systematically capturing the majority of the variance of the data, dPCA also exposes the dependence of the neural representation on task parameters such as stimuli, decisions, or rewards. To illustrate our method we reanalyze population data from four datasets comprising different species, different cortical areas and different experimental tasks. In each case, dPCA provides a concise way of visualizing the data that summarizes the task-dependent features of the population response in a single figure. DOI: http://dx.doi.org/10.7554/eLife.10989.001

Subfields: Artificial Intelligence, Cognitive Neuroscience

View Paper: https://www.semanticscholar.org/paper/79364486edaadaee4e6080c429a8f1a34987b8d1

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Artificial Intelligence

In Cognitive Neuroscience

Citations in the corpus

Computation Through Neural Population Dynamics (2020, methodology): “We start by discussing error-driven learning, which has a rich history in human psychophysics (for a review, see Krakauer et al. 2019).”
Complementary roles of dimensionality and population structure in neural computations (2020, methodology): “A common approach to characterizing the relationship between neural activity and the ongoing computation is to analyze the selectivity of individual cells by performing a linear regression of activity with respect to controlled or measured task variables, such as stimulus, context or behavioral choi”
Unsupervised Discovery of Demixed, Low-Dimensional Neural Dynamics across Multiple Timescales through Tensor Component Analysis (2017, methodology): “Data 769 were smoothed within each trial with a Gaussian filter with a standard deviation of 50 ms (same as in [34]).”
Long-term stability of cortical population dynamics underlying consistent behavior (2020, unknown): “Moreover, these latent dynamics exhibit some common characteristics across cortical regions 5,49 and even across species 16,50 .”
Cortical areas interact through a communication subspace (2019, methodology): “…whether our ability to predict V2 fluctuations involves only a subspace of V1 population activity, we used reduced-rank regression (Izenman, 1975; Kobak et al., 2016), a variant of linear regression in which the regression dimensions are constrained to lie in a low-dimensional subspace (see STAR…”
Representational models: A common framework for understanding encoding, pattern-component, and representational-similarity analysis (2017, methodology): “To 268 exploit the temporal detail provided by electrophysiological recordings, the analyses can 269 be either performed using a sliding window over the time course of the trial [40–42], or 270 by “stacking” the time series and conditions, resulting in a activity matrix with TK 271 rows [43].”
Artificial Neural Networks for Neuroscientists: A Primer (2020, methodology): “To facilitate side-by-side comparison between artiﬁcial and biological neural networks, activity of an ANN can be visualized and analyzed with the same dimensionality reduction tools (e.g., PCA) used for biological recordings [Mante et al., 2013, Kobak et al., 2016, Williams et al., 2018].”
Inferring single-trial neural population dynamics using sequential auto-encoders (2017, methodology): “Briefly, for each of the single-session models and the multi-session model, we performed demixing principal components analysis [20] on the factor outputs.”