- 博客(142)
- 资源 (5)
- 收藏
- 关注
原创 Matplotlib imshow() not show whole image
imshow() subplots with shared axes generate unwanted white spaces #1789
2016-01-13 14:35:53
2021
原创 sequence training
Sequence training is short for sequence discriminative training.ReferenceVoigtlaender, Paul, et al. “Sequence-discriminative training of recurrent neural networks.” Acoustics, Speech and Signal Proces
2016-01-06 23:09:51
1628
原创 configure with MKL
../configure CXXFLAGS="-DMKL_ILP64 -fopenmp -m64 -I/opt/intel/mkl/include" LDFLAGS="-Wl,--no-as-needed -L/opt/intel/mkl/lib/intel64" LDLIBS="-lmkl_intel_ilp64 -lmkl_core -lmkl_gnu_thread -ldl"Reference
2016-01-05 14:03:58
1693
原创 Imbalanced data
The Impact of Imbalanced Training Data for Convolutional Neural Networks Following this, oversampling is used on the imbalanced training sets to increase the performances to that of the balanced set.
2016-01-04 22:50:04
1540
原创 netCDF vs HDF
Re: netCDF vs HDFOpinions on NetCDF vs HDF5 for storing scientific data?
2016-01-03 13:34:48
2599
原创 OpenFST notes
symbol tableThe internal representation of an arc label is an integer.The symbols are used for I/O purpose.You can store the symbols table with the FST file or not. If you do, use the options --keep
2015-12-26 19:57:58
2015
原创 Thrax Tutorial
Thrax Tutorial Thrax is used for compiling grammars expressed as regular expressions and content-dependent rewrite rules into WFSTs.commandsthraxmakedep --save_symbols=truethraxcompiler -input_g
2015-12-26 15:49:41
2622
原创 Github logo in LaTeX
fontawesome – Font containing web-related icons The package offers access to the large number of web-related icons provided by the included font. The package requires the package, fontspec,
2015-12-24 12:26:18
1896
原创 kd-tree
papers:Silpa-Anan, Chanop, and Richard Hartley. “Optimised KD-trees for fast image descriptor matching.” CVPR 2008. IEEE Conference on. IEEE, 2008.
2015-12-21 22:06:05
1672
原创 kaldi parallelism
This post focus on run.pl of local parallelism.run.pl <options> <log-file> <command><options>: a job range specifier (e.g. JOB=1:4)<log-file>: file name for array jobs must contain identifier of the
2015-12-21 10:34:03
1515
原创 AWK programming
程序模型pattern { action }变量两种变量scalararray变量名内建变量大写全局变量首字母大写局部变量小写数组允许任意数字或字符串做索引内建变量 变量 说明 FILENAME input file name FNR 文件的record数 FS field分隔符 RS record分隔符 NF field数 NR
2015-12-20 14:37:35
1481
原创 Viterbi training
Viterbi training has another name segmental K-means.Within each state, a further alignment of observations to mixture components is made. HTK provides two methods for this: 1. use clustering to alloc
2015-12-20 00:08:09
1923
原创 visual vocabulary of binary features
How to use bag of words example with BRIEF descriptors?opencv c++ ORB Bag Of WordsHow to use bag of words example with BRIEF descriptors?Is there a way of using ORB with BOW?DBoW2: Enhanced hierarc
2015-12-19 00:20:30
1413
原创 kaldi increasing GMM components in the training procedure
This (continuous Gaussian distribution) allows simple single Gaussian distributions to be used for an initial untied model set where the training data is very patchy. Then once tying has been performed
2015-12-17 22:45:05
1377
转载 install Adobe Reader on Ubuntu
How to Install Adobe Reader in Ubuntu 14.10How do I install Adobe Acrobat Reader from the repository?
2015-12-17 17:46:32
1281
原创 WFST papers
M. Mohri, F. C. N. Pereira, and M. Riley. Weighted FiniteState Transducers in Speech Recognition. Computer Speech and Language, 16(1):69–88, 2002.Allauzen, C., Mohri, M., Riley, M., Roark, B., A Gener
2015-12-17 11:27:17
1542
转载 customize mac terminal prompt
Customize Your Shell & Command PromptHow to Customize the Command Prompt
2015-12-16 16:54:13
1820
转载 change default text editor to sublime text
How do I make Sublime Text 3 the default text editor [duplicate] Open /usr/share/applications/defaults.list , search for all instances of gedit.desktop and replace them with sublime_text.desktop. Sav
2015-12-16 14:50:49
1329
原创 Mentally grasp large project
How do you keep track of large projects? The trick to understanding a large code base is to not try to understand all of it. After a certain size, you can’t hold a mental model in your head of the en
2015-12-16 11:07:45
416
原创 python dict
LPTHW - Exercise 39: Dictionaries, Oh Lovely DictionariesPyCon 2010: The Mighty Dictionary
2015-12-15 10:02:27
340
原创 kaldi monophone training outline
This outline is based on the steps/train_mono.sh script.Use a subset of the training set to init the monophone modelgmm-init-mono [options] Topo-in Dim Model-out Tree-out gmm-init-mono "--train-feats=$
2015-12-10 22:29:24
1052
转载 mac install JDK
http://stackoverflow.com/questions/24342886/how-to-install-java-8-on-macDon’t rely on Oracle to install Java properly on your mac. Use HomeBrew.brew tap caskroom/caskbrew install brew-caskIf you g
2015-12-10 11:38:01
499
原创 whitening
Data matrix XX shape (D,N)(D, N). Zero mean centering or not depending on circumstance. Compute eigen vectors with SVD: U,S,V=linalg.svd(X)U, S, V = \texttt{linalg.svd}(X)Projection: X′=U⊺∗XX' = U
2015-12-08 21:43:02
634
原创 image pre-processing
For natural images, it make little sense to have separate mean and variance values for each pixel. They are already have equivalent variances.The desiderata are that our learning algorithms sees a tra
2015-12-08 16:04:31
558
原创 ASR evaluation
ASR evaluationTranscription accuracyWER(Word Error Rate)SER(Sentence Error Rate)Semantic accuracyConcept Accuracy(based on domain concepts)
2015-12-07 20:14:20
1347
原创 dot(.) is a synonym of `source`
In a bash script what does a dot followed by a space and then a path mean?It’s a synonym of the builtin source. It will execute commands from a file in the current shell, as read from help source or h
2015-12-04 23:04:38
446
原创 WFST over semiring
A weighted finite state transducer (WFST) T(x,y)T(x, y) over a semiring is defined as:T(x,y)=⊕π∈P(I,x,y,F)λ(p[π])⊗w[π]⊗ρ(n[π])T(x, y) = \underset{\pi\in P(I,x,y,F)}{\oplus}\lambda(p[\pi])\otimes w[\p
2015-12-03 17:03:49
932
原创 kaldi can link with openfst-1.5.0 without patch
openfst-1.5.0$ mkdir build && cd build$ ../configure --prefix=/opt/openfst-1.5.0 \--enable-static=no \--enable-far \--enable-ngram-fsts \--enable-pdt \--enable-python \LDLIBS="-ldl"kaldi-c09caf9
2015-12-02 19:22:02
922
原创 language model
ARPA format language modelhttp://www.statmt.org/book/slides/07-language-models.pdfToolkits: - SRILM - IRSTLM
2015-12-01 16:38:08
3434
1
原创 android jni
Compiling and Using a C++ Library on Android with Android Studio
2015-11-26 00:30:02
382
原创 Android Studio tips
place to put the so libraries:just put the platform directory in app/src/main/jniLibsJNI folder in Android StudioAndroid Studio Jar、so、library项目依赖Android Studio: Error:/android-studio/gradle/gradle-
2015-11-26 00:27:05
691
原创 IntelliJ IDEA font fix
Ugly fonts in Netbeans - How can i make it use the system font?How to fix font anti-aliasing in IntelliJ IDEA when using high DPI?
2015-11-24 22:47:34
712
原创 android development
What is Android SDK Build-tools and which version should be used?
2015-11-24 20:34:51
423
原创 bash"="后自动完成
Initially i thought this could be a bug but later realised that this behavior is because of readline treating ‘=’ as a word break character.# mkdir =foo =foobar# cd =\=foo-bash: cd: ==foo: No such fil
2015-11-20 10:43:32
449
原创 64bit Ubuntu supports 32bit libraries
Ubuntu 14.04 $ sudo apt-get install libc6:i386 libncurses5:i386 libstdc++6:i386 lib32z1Older version:$ sudo apt-get install ia32-libs
2015-11-10 11:29:23
603
原创 Caffe cuDNN max-pooling with in-place dropout
dropout in place incompatible with max pooling #117Fix Max pooling layer to use a mask #162Finish up max pooling with a mask from @sguada #448In-place computation can break gradient computation #201
2015-11-06 22:24:35
911
多尺度Retinex图像增强
2011-08-01
四六级词汇表,四六级词汇表
2011-07-20
空空如也
TA创建的收藏夹 TA关注的收藏夹
TA关注的人