Suse linux misc

Some system related tools like zypper, yast, ldconfig could not be accessed by usual users. Only when log in as root user can you find those commands.

To stopping the GUI and using the terminal on Ubuntu, do:

sudo /etc/init.d/gdm stop

then, press "Ctrl+Alt+F1" 

On the Suse server, to build the package Theano to work with Python, you have to compile the Python to shared library. That's to build Python in following steps:

./configure --enable-shared --prefix=xxxx

make 

sudo make install

Otherwise, it probably gives the error "libpython2.6.a: could not read symbols: Bad value"

Posted via email from Troy's posterous

Two more papers on speaker adaptation, 2010

Download now or preview on posterous

New Speaker Adaptation Method Using 2-D PCA.pdf (276 KB)

Download now or preview on posterous

Unsupervised Speaker Adaptation based on the Cosine Similarity for Text-Independent Speaker Verification.pdf (149 KB)

Posted via email from Troy's posterous

Two new speaker adaptation methods

Speaker adaptation using generalised low rank approximations of training matrices

http://scitation.aip.org/getabs/servlet/GetabsServlet?prog=normal&id=ELLEAK000046000010000724000001&idtype=cvips&gifs=yes&ref=no

Abstract:

A speaker adaptation method based on the low rank approximation of matrices (GLRAM) of training models is described. In the method, each model is represented as a matrix, and a set of such training matrices is decomposed into a set of speaker weights and two basis matrices for row and column spaces by reducing both row and column ranks of thetraining models. As a result, the speaker weight becomes amatrix, the row and column dimensions of which can beadjusted. In the isolated-word experiment, the proposed method showed better performance than both eigenvoice and MLLR for the adaptation data of about 20 s or longer.

Bilinear model for speaker adaptation using tensor analysis

http://scitation.aip.org/getabs/servlet/GetabsServlet?prog=normal&id=ELLEAK000046000003000243000001&idtype=cvips&gifs=yes&ref=no

Abstract:

A novel speaker adaptation method based on two-way analysis of training speakers is described. A set of training models is expressed as a tensor and is decomposed into two factors using nonlinear iterative partial least squares, producing a bilinear model. The resulting model has bases of lower dimension and more free parameters than those of eigenvoice, enabling more elaborate modelling for a moderate amount of adaptation data. Results from the isolated-word recognition test show that the proposed model outperforms both eigenvoice and maximum likelihood linear regression (MLLR) for adaptation data longer than 15 s. Moreover, the proposed method can straightforwardly be extended to n-way analysis, e.g. for simultaneous adaptation of speaker, environment, etc.

Posted via email from Troy's posterous

Mac keyboard shortcuts

http://www.danrodney.com/mac/index.html

Posted via email from Troy's posterous

Conditional RBM and Factored Conditional RBM

Download now or preview on posterous

Taylor_Graham_W_200911_PhD_thesis.pdf (10890 KB)

Posted via email from Troy's posterous

Two papers about mcRBM

Download now or preview on posterous

ranzato_aistats2010.pdf (2849 KB)

Download now or preview on posterous

ranzato_cvpr2010.pdf (2945 KB)

Posted via email from Troy's posterous

Install SGE on the server

1. Make sure there is no sge_qmaster and sge_execd running, if there are kill them.

2. Add a permanent ip to host name mapping in /etc/hosts:

    e.g. 172.23.34.54 speech5.ddns.comp.nus.edu.sg

2. Install the two packgaes:

    a. sun-sge-common-6.2-2.noarch.rpm

    b. sun-sge-bin-linux24-x64-6.2-2.x86_64.rpm

3. cd to the directory /gridware/sge

4. install qmaster using:

    ./install_qmaster

    most of the configurations are using default, do set the host to the same one as previous steps, e.g. speech5.ddns.comp.nus.edu.sg

5. install execd using:

    ./install_execd

   most of the configurations are set to default

then that's all!

Posted via email from Troy's posterous