Holy sh*t! We're in a paradigm shift.
In less than a decade, machine learning and scalable computing have demonstrated how knowledge can be extracted from complex datasets. The primary workhorse of this success has been free open source software (FOSS). With its strong emphasis on API design, the FOSS culture has made it less effortful to plan, develop in teams, re-use, distribute, teach, optimize & scale data analysis efforts. Coding sprints are a way to focus development efforts and share best practices that generalize across a range of application domains.
Supervised Neural Time Series is a week-long coding sprint to gather established data scientists, who specialize in high-dimensional neural time series. Together, we will work on advancing popular or upcoming FOSS projects that enable the analysis of a broad class of neural recordings: extracellular neurophysiology (spike trains), electro-corticography (ECoG) and magneto-/electro-encephalography (MEG/EEG).
- MNE is a single suite of tools for pre-processing, source-modeling, decoding and visualization of MEG, EEG and ECoG data.
- Pyglmnet enables generalized linear modeling with advanced regularization.
- Spykes offers 101 analysis tools for spike trains and visualization, including rasters, PSTHs, tuning curves, RF estimation and population decoding.
- PyRiemann offers tools to characterize the covariance structure of high-dimensional signals with Riemannian geometry.
We are looking for established data science developers to both contribute and bring their own projects. Although the organizers specialize in neural time series, we are looking to productively engage with data scientists from fields with similarly structure data (finance, musicology, speech processing, climate science, seismology, kinect and radar analyses, etc.). In keeping with this push for diversity, we particularly welcome women and underrepresented minorities.