Discovering event structure in continuous narrative perception and memory, bioRxiv, 2016-10-15
SummaryDuring realistic, continuous perception, humans automatically segment experiences into discrete events. Using a novel model of neural event dynamics, we investigate how cortical structures generate event representations during continuous narratives, and how these events are stored and retrieved from long-term memory. Our data-driven approach enables identification of event boundaries and event correspondences across datasets without human-generated stimulus annotations, and reveals that different regions segment narratives at different timescales. We also provide the first direct evidence that narrative event boundaries in high-order areas (overlapping the default mode network) trigger encoding processes in the hippocampus, and that this encoding activity predicts pattern reinstatement during recall. Finally, we demonstrate that these areas represent abstract, multimodal situation models, and show anticipatory event reinstatement as subjects listen to a familiar narrative. Our results provide strong evidence that brain activity is naturally structured into semantically meaningful events, which are stored in and retrieved from long-term memory.
biorxiv neuroscience 100-200-users 2016I Tried a Bunch of Things The Dangers of Unexpected Overfitting in Classification, bioRxiv, 2016-10-04
ABSTRACTMachine learning is a powerful set of techniques that has enhanced the abilities of neuroscientists to interpret information collected through EEG, fMRI, MEG, and PET data. With these new techniques come new dangers of overfitting that are not well understood by the neuroscience community. In this article, we use Support Vector Machine (SVM) classifiers, and genetic algorithms to demonstrate the ease by which overfitting can occur, despite the use of cross validation. We demonstrate that comparable and non-generalizable results can be obtained on informative and non-informative (i.e. random) data by iteratively modifying hyperparameters in seemingly innocuous ways. We recommend a number of techniques for limiting overfitting, such as lock boxes, blind analyses, and pre-registrations. These techniques, although uncommon in neuroscience applications, are common in many other fields that use machine learning, including computer science and physics. Adopting similar safeguards is critical for ensuring the robustness of machine-learning techniques.
biorxiv neuroscience 200-500-users 2016Empirical assessment of published effect sizes and power in the recent cognitive neuroscience and psychology literature, bioRxiv, 2016-08-26
AbstractWe have empirically assessed the distribution of published effect sizes and estimated power by extracting more than 100,000 statistical records from about 10,000 cognitive neuroscience and psychology papers published during the past 5 years. The reported median effect size was d=0.93 (inter-quartile range 0.64-1.46) for nominally statistically significant results and d=0.24 (0.11-0.42) for non-significant results. Median power to detect small, medium and large effects was 0.12, 0.44 and 0.73, reflecting no improvement through the past half-century. Power was lowest for cognitive neuroscience journals. 14% of papers reported some statistically significant results, although the respective F statistic and degrees of freedom proved that these were non-significant; p value errors positively correlated with journal impact factors. False report probability is likely to exceed 50% for the whole literature. In light of our findings the recently reported low replication success in psychology is realistic and worse performance may be expected for cognitive neuroscience.
biorxiv neuroscience 500+-users 2016Cytoarchitectonic similarity is a wiring principle of the human connectome, bioRxiv, 2016-08-07
AbstractUnderstanding the wiring diagram of the human cerebral cortex is a fundamental challenge in neuroscience. Elemental aspects of its organization remain elusive. Here we examine which structural traits of cortical regions, particularly their cytoarchitecture and thickness, relate to the existence and strength of inter-regional connections. We use the architecture data from the classic work of von Economo and Koskinas and state-of-the-art diffusion-based connectivity data from the Human Connectome Project. Our results reveal a prominent role of the cytoarchitectonic similarity of supragranular layers for predicting the existence and strength of connections. In contrast, cortical thickness similarity was not related to the existence or strength of connections. These results are in line with findings for non-human mammalian cerebral cortices, suggesting overarching wiring principles of the mammalian cerebral cortex. The results invite hypotheses about evolutionary conserved neurobiological mechanisms that give rise to the relation of cytoarchitecture and connectivity in the human cerebral cortex.
biorxiv neuroscience 100-200-users 2016AFNI and Clustering False Positive Rates Redux, bioRxiv, 2016-07-27
AbstractIn response to reports of inflated false positive rate (FPR) in FMRI group analysis tools, a series of replications, investigations, and software modifications were made to address this issue. While these investigations continue, significant progress has been made to adapt AFNI to fix such problems. Two separate lines of changes have been made. First, a long-tailed model for the spatial correlation of the FMRI noise characterized by autocorrelation function (ACF) was developed and implemented into the 3dClustSim tool for determining the cluster-size threshold to use for a given voxel-wise threshold. Second, the 3dttest++ program was modified to do randomization of the voxel-wise t-tests and then to feed those randomized t-statistic maps into 3dClustSim directly for cluster-size threshold determination-without any spatial model for the ACF. These approaches were tested with the Beijing subset of the FCON-1000 data collection. The first approach shows markedly improved (reduced) FPR, but in many cases is still above the nominal 5%. The second approach shows FPRs clustered tightly about 5% across all per-voxel p-value thresholds ≤ 0.01. If t-tests from a univariate GLM are adequate for the group analysis in question, the second approach is what the AFNI group currently recommends for thresholding. If more complex per-voxel statistical analyses are required (where permutationrandomization is impracticable), then our current recommendation is to use the new ACF modeling approach coupled with a per-voxel p-threshold of 0.001 or below. Simulations were also repeated with the now infamously “buggy” version of 3dClustSim the effect of the bug on FPRs was minimal (of order a few percent).
biorxiv neuroscience 0-100-users 2016Suite2p beyond 10,000 neurons with standard two-photon microscopy, bioRxiv, 2016-07-01
AbstractTwo-photon microscopy of calcium-dependent sensors has enabled unprecedented recordings from vast populations of neurons. While the sensors and microscopes have matured over several generations of development, computational methods to process the resulting movies remain inefficient and can give results that are hard to interpret. Here we introduce Suite2p a fast, accurate and complete pipeline that registers raw movies, detects active cells, extracts their calcium traces and infers their spike times. Suite2p runs on standard workstations, operates faster than real time, and recovers ~2 times more cells than the previous state-of-the-art method. Its low computational load allows routine detection of ~10,000 cells simultaneously with standard two-photon resonant-scanning microscopes. Recordings at this scale promise to reveal the fine structure of activity in large populations of neurons or large populations of subcellular structures such as synaptic boutons.
biorxiv neuroscience 200-500-users 2016