Fernando Pérez

Focal brain lesions to critical locations cause widespread disruption of the modular organization of the brain

Although it is generally assumed that brain damage predominantly affects only the function of the damaged region, here we show that focal damage to critical locations causes disruption of network organization throughout the brain. Using resting state fMRI, we assessed whole-brain network structure in patients with focal brain lesions. Only damage to those brain regions important for communication between subnetworks (e.g., “connectors”)—but not to those brain regions important for communication within sub-networks (e.g., “hubs”)—led to decreases in modularity, a measure of the integrity of network organization. Critically, this network dysfunction extended into the structurally intact hemisphere. Thus, focal brain damage can have a widespread, nonlocal impact on brain network organization when there is damage to regions important for the communication between networks. These findings fundamentally revise our understanding of the remote effects of focal brain damage and may explain numerous puzzling cases of functional deficits that are observed following brain injury.

Python: An Ecosystem for Scientific Computing

As the relationship between research and computing evolves, new tools are required to not only treat numerical problems, but also to solve various problems that involve large datasets in different formats, new algorithms, and computational systems such as databases and Internet servers. Python can help develop these computational research tools by providing a balance of clarity and flexibility without sacrificing performance.

Collaborative cloud-enabled tools allow rapid, reproducible biological insights

Microbial ecologists today face critical computational barriers. The rapid increase in the quantity of data acquired by modern sequencing instruments makes analysis by hand infeasible, and even software developed just a few years ago cannot scale to modern data sets. As a result, making advanced, scalable algorithms and large-scale computational resources available to end-users is necessary to advancing our understanding of microbial ecology.

Teaching computing with the IPython notebook (abstract only)

The IPython Notebook is an interactive browser-based environment where you can combine code execution, text, mathematics, plots, and rich media into a single document. Originally designed for use as an electronic lab notebook for computational science, it is increasingly being used in teaching as well, and a rich ecosystem of open source plugins and extensions for teaching is growing around it. The first half of this hands-on workshop will introduce the Notebook and present examples of lessons and instructional materials built around it. In the second half, attendees will explore future directions for the Notebook as a teaching platform. For more information, please view our GitHub repository online at https://github.com/gvwilson/sigcse2014-ipython-workshop.

Sparse reproducing kernels for modeling fiber crossings in diffusion weighted imaging

Using existing estimators found in High Angular Resolution Diffusion Imaging (HARDI), such as the Orientation Probability Distribution Function (OPDF) of Aganj et al., we develop a new mathematical framework for implementing HARDI. In contrast to traditional methods based on spherical harmonics, the framework is based on a reproducing kernel formalism combined with recently developed Gaussian-like quadratures for the sphere, which leads to an efficient and sparse representation for HARDI estimators. We demonstrate that the new framework results in reconstructions that are more robust to noise and have better angular resolution, compared to spherical harmonic based reconstructions.

A recurrent neural network for classification of unevenly sampled variable stars

Astronomical surveys of celestial sources produce streams of noisy time series measuring flux versus time (‘light curves’). Unlike in many other physical domains, however, large (and source-specific) temporal gaps in data arise naturally due to intranight cadence choices as well as diurnal and seasonal constraints1–5. With nightly observations of millions of variable stars and transients from upcoming surveys4,6, efficient and accurate discovery and classification techniques on noisy, irregularly sampled data must be employed with minimal human-in-the-loop involvement. Machine learning for inference tasks on such data traditionally requires the laborious hand-coding of domain-specific numerical summaries of raw data (‘features’)7. Here, we present a novel unsupervised autoencoding recurrent neural network8 that makes explicit use of sampling times and known heteroskedastic noise properties. When trained on optical variable star catalogues, this network produces supervised classification models that rival other best-in-class approaches. We find that autoencoded features learned in one time-domain survey perform nearly as well when applied to another survey. These networks can continue to learn from new unlabelled observations and may be used in other unsupervised tasks, such as forecasting and anomaly detection.A novel unsupervised autoencoding recurrent neural network produces state-of-the-art supervised classification models. This network can continue to learn from new unlabelled observations and may be used in other unsupervised tasks.