Serving the Quantitative Finance Community

Search found 41 matches

by OOglesby
December 3rd, 2013, 10:39 pm
Forum: Student Forum
Topic: Kalman filter stability problem
Replies: 1
Views: 5965

Kalman filter stability problem

<t>There can be several places and several reasons why a matrix can become singular. Unless the model used is truly linear, your process noise matrix, Q, should not be all zeros. For nonlinear models, having even small non-zero values on the diagonal of Q can help prevent singular matrices from form...
by OOglesby
September 17th, 2013, 4:02 pm
Forum: Programming and Software Forum
Topic: Is there an IDE for cygwin?
Replies: 12
Views: 10251

Is there an IDE for cygwin?

<t>If I remember correctly, to use Eclipse IDE with cygwin compilers you need to do the following steps:1) Download and uncompress Eclipse somewhere easily accessable within cygwin (i.e. /usr/local/eclipse)2) Start eclipse from the cygwin command line using something like /usr/llocal/eclipse/eclipse...
by OOglesby
September 7th, 2013, 8:17 pm
Forum: Programming and Software Forum
Topic: MATLAB : Reading txt-file with delimiter problem
Replies: 4
Views: 7170

MATLAB : Reading txt-file with delimiter problem

<t>JTB,Try the following code snipped and see if it can read your file. Textscan is a much more complicated function to use, but incredibly flexible. The text data will be stored in a cell array. There is plenty of documentation about cell arrays if you are not familiar with them.As a caveat, I have...
by OOglesby
April 12th, 2013, 8:28 am
Forum: Programming and Software Forum
Topic: C++ Matrix Libraries: What are the choices?
Replies: 15
Views: 10407

C++ Matrix Libraries: What are the choices?

As an aside for any MATLAB programmers out there, Eigen has syntax that allow performing an operation along all the rows or columns of a matrix just like MATLAB. Many of the vectorization tricks that are part of MATLAB transfer well to Eigen.
by OOglesby
March 20th, 2013, 9:01 pm
Forum: Numerical Methods Forum
Topic: Matlab ode45
Replies: 1
Views: 9210

Matlab ode45

<t>Normally, the odefun has the form: ydot = odefun(t, y). However, you can pass extra arguments into the function: ydot = odefun(t, y, arg1, arg2, argn). Below are two code examples. The first example passes in u and computes a, b, and c inside the function. The second example assumes u is constant...
by OOglesby
March 7th, 2013, 4:09 pm
Forum: Student Forum
Topic: Big O (Computational Complexity) in Matlab - What's going on?
Replies: 8
Views: 10408

Big O (Computational Complexity) in Matlab - What's going on?

<t>Function calls in MATLAB have a large overhead, so your implementation MUST be iterative and not recursive. One possibility is:1) Iteratively subdivide your data (just like the binary search)2) Stop the data subdivision when a block of data suitably sized. i.e. small enough the O(N) does not hurt...
by OOglesby
March 6th, 2013, 10:19 pm
Forum: Student Forum
Topic: How to edit/view built in Matlab code for FFT
Replies: 1
Views: 8830

How to edit/view built in Matlab code for FFT

<r>You do not have access to the code for MATLAB's FFT. If you remove the MATLAB specific wrapper functions, it uses the FFTW library under the hood (<URL url="http://www.fftw.org/">http://www.fftw.org/</URL>). MATLAB does provide a function that allows tune the runtime speed of fft. Type 'help fftw...
by OOglesby
September 22nd, 2012, 12:54 am
Forum: Programming and Software Forum
Topic: OpenCV
Replies: 7
Views: 12393

OpenCV

<t>OpenCV along with CUDA are some of the big libraries for image processing and computer vision. There are a few pieces that would be useful for finance, but I also think CUDA and true machine learning libraries are better suited.Browsing the documentation, there are various classifiers and cluster...
by OOglesby
August 29th, 2012, 11:25 pm
Forum: Programming and Software Forum
Topic: Do quants use high precision arithmetic in C++, or do they just stick to doubles?
Replies: 8
Views: 13030

Do quants use high precision arithmetic in C++, or do they just stick to doubles?

<t>As an engineer/programmer and not a quant, I have found that many cases sufficient accuracy can be obtained using doubles or even floats. Many accuracy problems can be solved be changing the order of operations, grouping different quantities together, or reversing loops to sum smallest to largest...
by OOglesby
July 4th, 2012, 9:09 pm
Forum: Technical Forum
Topic: Kalman Gain
Replies: 8
Views: 17748

Kalman Gain

<t>Quotestate matrix (\Phi ) has complex eigenvaluesAll my experience with physics-based models. I am sure the eigenvalues related to the stability of the system, but I do not know how.QuoteAlso the model does not work so well with high dimensions (22) but better with lower dimensions (7)In my exper...
by OOglesby
July 4th, 2012, 6:43 pm
Forum: Technical Forum
Topic: Kalman Gain
Replies: 8
Views: 17748

Kalman Gain

<t>Q effectively sets a floor and the covariance is not allowed to shrink smaller than that floor. Q is also a trade-off between the smallest allowable covariance and the amount of smoothing that occurs. Therefore, a large Q will give a larger covariance at convergence, but will perform less smoothi...
by OOglesby
July 4th, 2012, 4:52 am
Forum: Technical Forum
Topic: Kalman Gain
Replies: 8
Views: 17748

Kalman Gain

<t>For a simple linear Kalman filter, I do not think the magnitude of the Kalman gain should ever be < 0. For reference the Kalman filter uses the following equations:The Kalman gain is the ratio of two positive quantities. A negative gain can result if two conditions are met:1. The correction equat...
by OOglesby
June 6th, 2012, 11:00 pm
Forum: Student Forum
Topic: Kalman Filter
Replies: 2
Views: 13544

Kalman Filter

<t>I give you my thoughts on your questions with the caveat that I have used Kalman filters in engineering but not finance. Additionally, I rarely use linear Kalman filters, but Extended Kalman filters.Quote1) A little bit of fat tails? If the noise(s) are not Gaussian, but e.g. Student-T; how much ...
by OOglesby
March 2nd, 2012, 6:35 pm
Forum: Technical Forum
Topic: Buy-Sell-Hold classification with Neural network
Replies: 22
Views: 19101

Buy-Sell-Hold classification with Neural network

<t>Classifiers such as neural networks or support vector machines are not magic. I feel they both require the user to be very knowledgeable about the problem domain.QuoteI am using 50 technical indicators as input and target vector is buy-sell-hold values.I would say that many of your inputs do not ...
by OOglesby
December 18th, 2011, 8:09 am
Forum: The Quantitative Finance Code Library Project
Topic: Choice of matrix library and interface
Replies: 95
Views: 35690

Choice of matrix library and interface

<t>Dmitry,Most of my comments about PETSc were in the context of financial engineering and development. The primary purpose of my review was a survey of general purpose vector/matrix data types.I am actually very impressed with PETSc's capabilities. My career consists of nonlinear optimization, diff...