Birds-of-a-feather sessions
We usually hold Birds of a Feather sessions in the evenings. This is a great occasion to discuss issues directly with the community, rather than on a mailing list.
Current schedule and locations for the BoFs can be found on the schedule on the conference site.
Proposed BoF sessions
Please add/vote for topics below!
- Topic (coordinator: Person)
- Attendee 1
- Attendee 2
- Partial Differential Equations (coordinator: ?) (proposed by Chris Kees)
- Chris Kees
- David Goldsmith
- Jon Guyer
Astronomy (coordinator: James Turner)
Meet at 19:30 on Thursday on the steps in front of the (Beckman Institute) lecture hall and we'll probably head off somewhere for more drinks/food following the reception.
- James Turner
- Christopher Hanley
- Perry Greenfield
- Joe Harrington (can attend 5pm Wed - 8pm Fri)
- Craig Allen
- Machine Learning/Probabilistic Modeling (coordinator: David Warde-Farley)
- David Warde-Farley
- Gael Varoquaux
- Fernando Perez
- C Coders (coordinator: ?) (proposed by Charles R. Harris)
- Charles R Harris
- Documentation (coordinator: David Goldsmith?)
- David Goldsmith
- Pauli Virtanen
- Joe Harrington (can attend 5pm Wed - 8pm Fri)
- Cython (coordinator: Dag Sverre Seljebotn)
- Darren Dale
- Kurt Smith
- Dag Sverre Seljebotn
- Christopher Barker
- David Warde-Farley
- Organization, Funding, and Future Direction of SciPy (coordinator: Joe Harrington, sergeant-at-arms: David Goldsmith)
- Joe Harrington (can attend 5pm Wed - 8pm Fri)
- David Goldsmith
- Behavioral Science and Data Wrangling (suggester: Dav Clark)
- Dav Clark
- Andrew Straw
- Gökhan Sever
Partial Differential Equations
List possible topics/areas of interest here.
Astronomy
Co-ordinating our development of astronomy libraries and whatever else people bring up.
Machine Learning/Probabilistic Modeling
What I (David W-F) am interested in is primarily reusable tools for doing {maximum likelihood, Bayesian} probabilistic modeling in Python. Lately a lot of options have popped up for quickly specifying and fitting/simulating models, including:
- PyMC 2.0 - http://pymc.googlecode.com/
- Infer.NET - http://research.microsoft.com/en-us/um/cambridge/projects/infernet/
- Hierarchical Bayesian Compiler - http://www.cs.utah.edu/~hal/HBC/
Obviously PyMC is the most promising given our preferred programming language but the others include interesting ideas too, for example Infer.NET features a compilation step whereby models are compiled to fast machine code; HBC can do the same by generating C code.
Some possible topics might include:
- The learn scikit - what's it good at, what's it missing?
- The maxentropy module - it took me a while to stumble upon it and recognize it as what I know as exponential family models. Could this code be reused, expanded, made more flexible?
- The old idea of porting Kevin Murphy's Bayes net toolbox, which DavidC got permission to incorporate under the BSD license but none of us actually got around to doing. :) [Edit: there is a Python toolbox available at http://dip.sun.ac.za/vision/trac-git/agouws-GrMPy.bzr]
- Things that make complicated models easier to implement and test, i.e. automatic differentiation (Sebastian Walter's pyadolc project is really cool and looks promising!)
- New algorithms that might deserve a place in scipy.cluster, ways of improving the existing implementations
- Package interoperability, data formats
- Accelerating adoption among the machine learning community who are by and large MATLAB users
- ??? (your idea here)
C Coders
List possible topics/areas of interest here.
Cython
Some topics:
- Move towards PEP3118 as the primary array protocol rather than ndarray?
- Cython templates?
- Native SIMD in Cython -- good or bad?
- Future direction of Cython in general
Documentation
Topics to discuss:
- Wrapping up NumPy docstrings: review process
- scipy package docs
- Tutorials
- The SciPy User Guide: TOC, editorial process, timing
- Organizing/indexing the Cookbook
Organization, Funding, and Future Direction of SciPy
I'd like to spend a strict 10 minutes on each of these, cutting off discussion and moving on after each item. After all 6 items, we can continue discussion on any item:
- What are our long-term goals?
- What are our current strengths and weaknesses?
- How is our current Steering Committee/grass-roots model working?
- Would funding help? What would we do with it? Would it require a change in how the community operates?
- How can we get funding?
- In the large, how should we proceed?
Behavioral Science and Data Wrangling
- Overview of programming with and for beginning programmer / scientists (from Euro-CNS 2009)
- Approaches to data storage / management / visualization (i.e. w/ netCDF or HDF5)
- Stimulus presentation / Response collection (or more general experimental issues)
- Hardware interfaces for I/O (LabJack, arduino?)