Computer simulations are invaluable tools for scientific discovery. However, accurate simulations are often slow to execute, which limits their applicability to extensive parameter exploration, large-scale data analysis, and uncertainty quantification. A promising route to accelerate simulations by building fast emulators with machine learning requires large training datasets, which can be prohibitively expensive to obtain with slow simulations. Here we present a method based on neural architecture search to build accurate emulators even with a limited number of training data. The method successfully accelerates simulations by up to 2 billion times in 10 scientific cases including astrophysics, climate science, biogeochemistry, high energy density physics, fusion energy, and seismology, using the same super-architecture, algorithm, and hyperparameters. Our approach also inherently provides emulator uncertainty estimation, adding further confidence in their use. We anticipate this work will accelerate research involving expensive simulations, allow more extensive parameters exploration, and enable new, previously unfeasible computational discovery.

Python haters always say, that one of reasons they don't want to use it, is that it's slow. Well, whether specific program - regardless of programming language used - is fast or slow is very much dependant on developer who wrote it and their skill and ability to write optimized and fast programs.

So, let's prove some people wrong and let's see how we can improve performance of our Python programs and make them really fast!

We wish to take an array of 32 million 64bit floating point values, and compute the sum of their squares. This will let us explore some fundamental abilities of various languages. Their ability to iterate over arrays efficiently, whether they can vectorize basic loops, and whether higher order functions like map and reduce compile to efficient code. When applicable, I will show runtimes of both map and reduce, so we get insight into whether the language can stream higher order functions together, and also the runtime with a single reduce or fold operation.

elkai is a Python 3 library for solving travelling salesman problems without external dependencies, based on LKH by Keld Helsgaun.

@INPROCEEDINGS{7170976,
author={Risbeck, Michael J. and Maravelias, Christos T. and Rawlings, James B. and Turney, Robert D.},
booktitle={American Control Conference (ACC), 2015},
title={Cost optimization of combined building heating/cooling equipment via mixed-integer linear programming},
year={2015},
month={July},
pages={1689-1694},
abstract={In this paper, we propose a mixed-integer linear program to economically optimize equipment usage in a central heating/cooling plant subject to time-of-use and demand charges for utilities. The optimization makes both discrete on/off and continuous load decisions for equipment while determining utilization of thermal energy storage systems. This formulation allows simultaneous optimization of heating and cooling subsystems, which interact directly when heatrecovery chillers are present. Nonlinear equipment models are approximated as piecewise-linear to balance modeling accuracy with the computational constraints imposed by online implementation and to ensure global optimality for the computed solutions. The chief benefits of this formulation are its ability to tightly control on/off switching of equipment, its consideration of cost contributions from auxiliary equipment such as pumps, and its applicability to large systems with multiple heating and cooling units in which a combinatorial problem must be solved to pick the optimal mix of equipment. These features result in improved performance over heuristic scheduling rules or other formulations that do not consider discrete decision variables. We show optimization results for a system with four conventional chillers, two heat-recovery chillers, and one hot water boiler. With a timestep of 1 h and a horizon of 48 h, the optimization problem can be solved to optimality within 5 minutes, indicating suitability for online implementation.},
keywords={Biological system modeling;Cooling;Generators;Load modeling;Optimization;Production;Switches},
doi={10.1109/ACC.2015.7170976},}