zi-w / max-value-entropy-search Goto Github PK
View Code? Open in Web Editor NEWMax-value Entropy Search for Efficient Bayesian Optimization
License: MIT License
Max-value Entropy Search for Efficient Bayesian Optimization
License: MIT License
Dear @zi-w ,
While checking on your nice paper about MES I noticed that Shekel is introduced with 10 dimensions in the text but looking at the code here and the function definition here it appears to be 4-dimensional.
Could you confirm that the plots of the paper correspond to 4-dimensional inputs?
Thank you in advance for checking this comment.
Best regards.
Max-value-Entropy-Search/utils/sampleMaximumValues.m
Lines 31 to 32 in 8601220
Dear Dr. Wang,
I see this script is adapted from the code for the paper PES, and these two lines are not modified. I see that b
is from a uniform distribution in (0, 2*pi), however, I do not understand how this W
comes from? why can we element-wise multiply the square root of l
the length scale ?
Thank you for your time and help in advance.
Question 1:
In Synthetic Functions, how to calculate the max f(x) which is the first term of simple and inference regret?
Question 2:
In Optimization Test Functions, what are the input domains of three test functions?
What's the use for this line? D
itself is already the eigendecomposition of Z^t Z + \sigma^2 I
, it looks to me that R=1./D
. Can you please clarify this ?
Dear Dr. Wang,
I found your paper about MES really interesting. I have a question about it: since I'm working on discrete and scattered search spaces with "real" black box functions, is your code suitable for this as it is or do you suggest to adapt it? In particular, I refer to all the optimization scripts: can I avoid all the gradient-based searches with fmincon, just computing effectively the acquisition function in all the domain points?
Thank you in advance.
Dear Dr. Wang,
I'm considering using your 14-dimensional robot-pushing problem as a benchmark for my ongoing work on (heterogeneous-cost high-dimensional) Bayesian optimization.
While reviewing your awesome paper and codes, I found generate_simudata_2robot2thing.py provided 14 input parameters without their meaning and range:
rx = float(sys.argv[1])
ry = float(sys.argv[2])
xvel = float(sys.argv[3])
yvel = float(sys.argv[4])
simu_steps = int(float(sys.argv[5]) * 10)
init_angle = float(sys.argv[6])
rx2 = float(sys.argv[7])
ry2 = float(sys.argv[8])
xvel2 = float(sys.argv[9])
yvel2 = float(sys.argv[10])
simu_steps2 = int(float(sys.argv[11]) * 10)
init_angle2 = float(sys.argv[12])
rtor = float(sys.argv[13])
rtor2 = float(sys.argv[14])
According to your comment in robot_pushing_4.m:
% tuning range of x:
% xmin = [-5; -5; 1; 0];
% xmax = [5; 5; 30; 2*pi];
the range of robot location, pushing duration time, and the initial angle should be [-5, 5], [1, 30], and [0, 2*pi]. How about the other parameters? I presume xvel and yvel are velocities, and rtor is torque, but I have no idea what their range should be. Could you provide clarification on their meanings and range?
Thanks! I look forward to any information you can share.
Hello,
is it possible, that you can upload a mex file for windows for chol2invchol.c.
I tried several ways to mex it by myself, but I was not successful.
Thanks in advance.
Roman
Just wondering if the code can be used with open source products?
Best,
Andrew
In the section 5.2, do you ask for the maximum or minimum of the three function(eggholder, shekel, michalewicz)?
A declarative, efficient, and flexible JavaScript library for building user interfaces.
๐ Vue.js is a progressive, incrementally-adoptable JavaScript framework for building UI on the web.
TypeScript is a superset of JavaScript that compiles to clean JavaScript output.
An Open Source Machine Learning Framework for Everyone
The Web framework for perfectionists with deadlines.
A PHP framework for web artisans
Bring data to life with SVG, Canvas and HTML. ๐๐๐
JavaScript (JS) is a lightweight interpreted programming language with first-class functions.
Some thing interesting about web. New door for the world.
A server is a program made to process requests and deliver data to clients.
Machine learning is a way of modeling and interpreting data that allows a piece of software to respond intelligently.
Some thing interesting about visualization, use data art
Some thing interesting about game, make everyone happy.
We are working to build community through open source technology. NB: members must have two-factor auth.
Open source projects and samples from Microsoft.
Google โค๏ธ Open Source for everyone.
Alibaba Open Source for everyone
Data-Driven Documents codes.
China tencent open source team.