Machine Learning
visionfans
研究方向:计算机视觉,模式识别,机器学习
展开
-
Nomial
Nomial原创 2015-04-04 22:24:24 · 1575 阅读 · 0 评论 -
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原创 2015-12-08 16:04:31 · 562 阅读 · 0 评论 -
ASR evaluation
ASR evaluationTranscription accuracyWER(Word Error Rate)SER(Sentence Error Rate)Semantic accuracyConcept Accuracy(based on domain concepts)原创 2015-12-07 20:14:20 · 1358 阅读 · 0 评论 -
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原创 2016-01-06 23:09:51 · 1638 阅读 · 0 评论 -
language model
ARPA format language modelhttp://www.statmt.org/book/slides/07-language-models.pdfToolkits: - SRILM - IRSTLM原创 2015-12-01 16:38:08 · 3450 阅读 · 1 评论 -
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原创 2015-12-03 17:03:49 · 938 阅读 · 0 评论 -
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.原创 2016-01-04 22:50:04 · 1544 阅读 · 0 评论 -
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原创 2015-12-20 00:08:09 · 1931 阅读 · 0 评论 -
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原创 2015-12-17 11:27:17 · 1549 阅读 · 0 评论 -
HMM training
2 road lines: - EM algorithm(Baum-Welch) - Gradient DescentThe main advantage of the Baum-Welch algorithm (and hence the ML training) is due to its simplicity and the fact that requires no parameter原创 2015-04-26 20:12:19 · 694 阅读 · 0 评论 -
information retrieval
Key concepts:tf-idfinverted indexReferencesHow a Search Engine Works Information Retrieval: What is inverted index?Implementing Search Engines原创 2015-10-04 10:31:17 · 648 阅读 · 0 评论 -
loss functions for NN
cross entropy or log lossIn binary classification, the output is a single probability value yy. L(x,z)=−(zln(y)+(1−z)ln(1−y))\mathcal{L}(x, z)=-(z\ln(y)+(1-z)\ln(1-y))For multi-class problem, the ou原创 2015-09-11 16:54:08 · 569 阅读 · 0 评论 -
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原创 2015-12-08 21:43:02 · 644 阅读 · 0 评论