The Society for Computational
Economics will sponsor two optional pre-conference
workshops for conference attendees. Registration for workshops is available online with conference
registration. The fee is $25 for each full-day workshop. Morning and afternoon coffee breaks
are included; attendees will be on their own
for lunch. Space is limited and priority will be given to graduate
students and recent PhDs (earned November 1, 2013 or later). Non-students may register but will be waitlisted until after
April 1, 2014. Those who register and cannot be accommodated will be
refunded in full. Both workshops will be held on Saturday, 21 June at BI
Norwegian Business School.
As of May 20, 2014 everyone who registered for the workshops is confirmed,
and space is open for additional attendees in both workshops.
Scientific Computing in Python and Julia
Presented by
John Stachurski, Sébastien Villemot and Pablo Winant
Making
the right choice of programming language is central to the success of a
scientific computing project. Despite their important past achievements,
languages like Fortran, C or MATLAB suffer from limitations compared to
newly introduced languages. The goal of this workshop is therefore to
familiarize the audience with two recent languages, Python and Julia, that
offer unique improvements in the trade-off between expressiveness and
performance. The workshop will also highlight the use of these languages
in economics and finance.
Thanks to its user-friendliness and flexibility, and with the maturation
of a complete scientific stack of software, the Python language is
becoming increasingly popular in the industry and in many academic fields.
Recent development of compilation techniques have also made it a valid
contender for writing high performance computing, without requiring a lot
of low-level programming.
More recently, and with a more specific focus on scientific computing, a
team of researcher created the Julia language that offers modern syntax
and expressiveness, high performance (close to compiled languages) and an
easy transition from MATLAB. Both languages are free and open-source
software. They also have a lively scientific community behind them.
Attendees are expected to bring their laptop. Rudimentary knowledge of
programming concepts would be useful but by no way required. The workshop
will consist in three hands-on sessions:
- Morning session (John Stachuchski): Python in Economics and
Finance
General introduction to Python (syntax, datatypes, ...), and its
scientific libraries (scipy, matplotlib, statsmodels, pandas)
- Afternoon session 1 (Pablo Winant): High Performance Computing
with Python
Profiling and accelerating Python code by interfacing C libraries (cffi)
or by compiling Python code (Cython or numba)
- Afternoon session 2 (Sébastien Villemot): The Julia Language: a
fresh approach to scientific computing
Introduction to Julia: syntax, performance, ecosystem, comparison with
MATLAB, solving DSGE models
Numerical
Methods for Large Scale Dynamic Economic Models
Presented by
Lilia Maliar, Stanford University and Serguei Maliar, Santa
Clara University
Lilia
Maliar and Serguei Maliar will teach a one-day workshop about numerical
methods for solving large scale dynamic economic models. Applications
include large-scale models of international trade and a medium-scale new
Keynesian model. MATLAB codes will be provided to the workshop
participants.
Materials. Lilia Maliar and Serguei Maliar, (2013). “Numerical Methods
for Large Scale Dynamic Economic Models” in: Schmedders, K. and K. Judd
(Eds.),
Handbook of Computational Economics,
Volume 3, Amsterdam: Elsevier Science (forthcoming).
Abstract. We survey numerical methods that can accurately and reliably
solve dynamic economic models with 10-100 state variables using a standard
desktop computer and serial Matlab software. (Examples of such models are
new Keynesian models, life-cycle models, heterogeneous agents models, asset
pricing models, multisector models, multicountry models, and climate change
models.) First, we show efficient nonproduct techniques for interpolating
and approximating functions (Smolyak, stochastic simulation, and
ε-distinguishable set grids), accurate low-cost monomial integration
formulas, derivative-free solvers, and numerically stable regression
methods. Second, we discuss endogenous grid and envelope condition methods
that reduce the cost and increase accuracy of value function iteration.
Third, we show precomputation techniques that construct solution manifolds
for some models' variables outside the main iterative cycle. Fourth, we
review techniques that increase the accuracy of perturbation methods: a
change of variables and a hybrid of local and global solutions. Finally, we
show examples of parallel computation using multiple CPUs and GPUs including
applications on a supercomputer. We illustrate the performance of the
surveyed methods using a multi-agent model and a new Keynesian model. Matlab
codes will be made available to the workshop participants.
Contents:
1.
Nonproduct approaches to representing, approximating and interpolating
functions (Smolyak, stochastic simulation, and ε-distinguishable set grids).
2.
Nonproduct approaches to approximation of integrals (monomial formulas,
Monte Carlo and non-parametric methods).
3.
Derivative-free optimization methods.
4.
Dynamic programming methods for high-dimensional problems (endogenous grid
method and envelope condition method).
5.
Precomputation techniques (precomputation of integrals and precomputation of
intratemporal choice functions).
6.
Increasing accuracy of local (perturbation) methods (techniques of a change
of variables and a hybrid of global and local solution methods).
7.
Parallel computation on desktop and supercomputers.
8.
Numerical examples: multi-agent model and a new Keynesian model.
Prof.
Lilia Maliar and Serguei Maliar both obtained their PhD degrees in 1999 from
Universitat Pompeu Fabra, Spain. They taught at Stanford University and
University of Alicante; Serguei also taught at Leavey School of Business,
Santa Clara University. Their joint research on numerical methods for
dynamic economic models has appeared in numerous economic journals including
Computational Economics, Journal of
Business and Economic Statistics, Journal of Economic Dynamic and Control,
Journal of Money
Banking and Credit, Quantitative Economics,
Review of Economic Dynamics;
they contributed a chapter to
Handbook of Computational Economics;
and are currently advising Bank of Canada on the model for the optimal
monetary choice. |
|
BI Norwegian Business School |