GNU Octave is a high-level language, primarily intended for numerical
computations. It provides a convenient command line interface for
solving linear and nonlinear problems numerically.
The tca package is a Matlab program that implements the tree-dependent
component analysis (TCA) algorithms that extends the independent
component analysis (ICA), where instead of looking for a linear transform
that makes the data components independent, we are looking for components
that can be best fitted in a tree structured graphical model. The TCA model
can be applied in any situation where the data can be assumed to have been
transformed by an unknown linear transformation.
The main features of the considered identification problem are that there is no an a priori separation of the variables into inputs and outputs and the approximation criterion, called misfit, does not depend on the model representation. The misfit is defined as the minimum of the l2-norm between the given time series and a time series that is consistent with the approximate model. The misfit is equal to zero if and only if the model is exact and the smaller the misfit is (by definition) the more accurate the model is. The considered model class consists of all linear time-invariant systems of bounded complexity and the complexity is specified by the number of inputs and the smallest number of lags in a difference equation representation. We present a Matlab function for approximate identification based on misfit minimization. Although the problem formulation is representation independent, we use input/state/output representations of the system in order
英文版G.729語音壓縮標準。
GENERAL ASPECTS OF DIGITAL TRANSMISSION
SYSTEMS
CODING OF SPEECH AT 8 kbit/s
USING CONJUGATE-STRUCTURE
ALGEBRAIC-CODE-EXCITED
linear-PREDICTION (CS-ACELP)
The inverse of the gradient function. I ve provided versions that work on 1-d vectors, or 2-d or 3-d arrays. In the 1-d case I offer 5 different methods, from cumtrapz, and an integrated cubic spline, plus several finite difference methods.
In higher dimensions, only a finite difference/linear algebra solution is provided, but it is fully vectorized and fully sparse in its approach. In 2-d and 3-d, if the gradients are inconsistent, then a least squares solution is generated
MATLAB Code for Optimal Quincunx Filter
Bank Design
Yi Chen
July 17, 2006
This file introduces the MATLAB code that implements the two algorithms (i.e., Algorithms
1 and 2 in [1], or Algorithms 4.1 and 4.2 in [2]) used for the construction of
quincunx filter banks with perfect reconstruction, linear phase, high coding gain, certain
vanishing moments properties, and good frequency selectivity. The code can be
used to design quincunx filter banks with two, three, or four lifting steps. The SeDuMi
Matlab toolbox [3] is used to solve the second-order cone programming subproblems
in the two algorithms, and must be installed in order for this code to work.
measure through
the cross-entropy of test data. In addition,
we introduce two novel smoothing techniques,
one a variation of Jelinek-Mercer
smoothing and one a very simple linear interpolation
technique, both of which outperform
existing methods.
Batch version of the back-propagation algorithm.
％ Given a set of corresponding input-output pairs and an initial network
％ [W1,W2,critvec,iter]=batbp(NetDef,W1,W2,PHI,Y,trparms) trains the
％ network with backpropagation.
％
％ The activation functions must be either linear or tanh. The network
％ architecture is defined by the matrix NetDef consisting of two
％ rows. The first row specifies the hidden layer while the second
％ specifies the output layer.
％
％ Train a two layer neural network with the Levenberg-Marquardt
％ method.
％
％ If desired, it is possible to use regularization by
％ weight decay. Also pruned (ie. not fully connected) networks can
％ be trained.
％
％ Given a set of corresponding input-output pairs and an initial
％ network,
％ [W1,W2,critvec,iteration,lambda]=marq(NetDef,W1,W2,PHI,Y,trparms)
％ trains the network with the Levenberg-Marquardt method.
％
％ The activation functions can be either linear or tanh. The
％ network architecture is defined by the matrix NetDef which
％ has two rows. The first row specifies the hidden layer and the
％ second row specifies the output layer.
