Justin Kiggins

Product. Data. Science.

Managing a neuroscience data pipeline with airflow

03 April 2016

A few months ago, I found myself with a backlog of data that needed preprocessing before I could start really analyzing it. I decided I need some kind of a scheduler to help push my data through the preprocessing efficiently.

Read More

Generating Channel Graphs for KlustaKwik Automatically

03 July 2015

In order to build the masks for KlustaKwik, SpikeDetekt2 needs to know the graph of adjacencies between all of the recording sites on a recording array. However, a bad channel for a recording session can drastically change the requirements for a probe geometry.

Read More

Defining Large Array Geometries for Klustakwik

02 July 2015

We’ve started using KlustaKwik for spike sorting in the lab and I’ve been working on getting a data pipeline going. It’s masked EM algorithm is very clever, taking into account the adjacency matrix of an electrode array to determine whether it should use one, two, or n recording sites when clustering.

Read More