About: A community detection method based on constrained fractional set programming (CFSP). Changes:Initial Announcement on mloss.org.

About: Toeblitz is a MATLAB/Octave package for operations on positive definite Toeplitz matrices. It can solve Toeplitz systems Tx = b in O(n*log(n)) time and O(n) memory, compute matrix inverses T^(1) (with free log determinant) in O(n^2) time and memory, compute log determinants (without inverses) in O(n^2) time and O(n) memory, and compute traces of products A*T for any matrix A, in minimal O(n^2) time and memory. Changes:Adding a writeup in written/toeblitz.pdf describing the package.

About: A generalized version of spectral clustering using the graph pLaplacian. Changes:

About: Crino: a neuralnetwork library based on Theano Changes:1.0.0 (7 july 2014) :  Initial release of crino  Implements a torchlike library to build artificial neural networks (ANN)  Provides standard implementations for : * autoencoders * multilayer perceptrons (MLP) * deep neural networks (DNN) * input output deep architecture (IODA)  Provides a batchgradient backpropagation algorithm, with adaptative learning rate

About: This package is an implementation of a linear RankSVM solver with nonconvex regularization. Changes:Initial Announcement on mloss.org.

About: MSVMpack is a Multiclass Support Vector Machine (MSVM) package. It is dedicated to SVMs which can handle more than two classes without relying on decomposition methods and implements the four MSVM models from the literature: Weston and Watkins MSVM, Crammer and Singer MSVM, Lee, Lin and Wahba MSVM, and the MSVM2 of Guermeur and Monfrini. Changes:

About: GradMC is an algorithm for MR motion artifact removal implemented in Matlab Changes:Added support for multirigid motion correction.

About: minFunc is a Matlab function for unconstrained optimization of differentiable realvalued multivariate functions using linesearch methods. It uses an interface very similar to the Matlab Optimization Toolbox function fminunc, and can be called as a replacement for this function. On many problems, minFunc requires fewer function evaluations to converge than fminunc (or minimize.m). Further it can optimize problems with a much larger number of variables (fminunc is restricted to several thousand variables), and uses a line search that is robust to several common function pathologies. Changes:Initial Announcement on mloss.org.

About: The glmie toolbox contains scalable estimation routines for GLMs (generalised linear models) and SLMs (sparse linear models) as well as an implementation of a scalable convex variational Bayesian inference relaxation. We designed the glmie package to be simple, generic and easily expansible. Most of the code is written in Matlab including some MEX files. The code is fully compatible to both Matlab 7.x and GNU Octave 3.2.x. Probabilistic classification, sparse linear modelling and logistic regression are covered in a common algorithmical framework allowing for both MAP estimation and approximate Bayesian inference. Changes:added factorial mean field inference as a third algorithm complementing expectation propagation and variational Bayes generalised nonGaussian potentials so that affine instead of linear functions of the latent variables can be used

About: A Matlab implementation of Multilinear PCA (MPCA) and MPCA+LDA for dimensionality reduction of tensor data with sample code on gait recognition Changes:

About: Stochastic neighbor embedding originally aims at the reconstruction of given distance relations in a lowdimensional Euclidean space. This can be regarded as general approach to multidimensional scaling, but the reconstruction is based on the definition of input (and output) neighborhood probability alone. The present implementation also allows for handling dissimilarity or scoreinduced neighborhood topologies and makes use of quasi 2nd order gradientbased (l)BFGS optimization. Changes:

About: The aim is to embed a given data relationship matrix into a lowdimensional Euclidean space such that the point distances / distance ranks correlate best with the original input relationships. Input relationships may be given as (sparse) (asymmetric) distance, dissimilarity, or (negative!) score matrices. Inputoutput relations are modeled as lowconditioned. (Weighted) Pearson and soft Spearman rank correlation, and unweighted soft Kendall correlation are supported correlation measures for input/output object neighborhood relationships. Changes:

About: Regularization paTH for LASSO problem (thalasso) thalasso solves problems of the following form: minimize 1/2X*betay^2 + lambda*sumbeta_i, where X and y are problem data and beta and lambda are variables. Changes:Initial Announcement on mloss.org.

About: This toolbox implements a novel visualization technique called Sectors on Sectors (SonS), and a extended version called Multidimensional Sectors on Sectors (MDSonS), for improving the interpretation of several data mining algorithms. The MDSonS method makes use of Multidimensional Scaling (MDS) to solve the main drawback of the previous method, namely, the lack of representing distances between pairs of clusters. These methods have been applied for visualizing the results of hierarchical clustering, Growing Hierarchical SelfOrganizing Maps (GHSOM), classification trees and several manifolds. These methods make possible to extract all the existing relationships among centroids’ attributes at any hierarchy level. Changes:Initial Announcement on mloss.org.

About: A collection of python code to perform research in optimization. The aim is to provide reusable components that can be quickly applied to machine learning problems. Used in:  Ellipsoidal multiple instance learning  difference of convex functions algorithms for sparse classfication  Contextual bandits upper confidence bound algorithm (using GP)  learning output kernels, that is kernels between the labels of a classifier. Changes:

About: Soltion developed by team Turtle Tamers in the ChaLearn Gesture Challenge (http://www.kaggle.com/c/GestureChallenge2) Changes:Initial Announcement on mloss.org.

About: The VLFeat open source library implements popular computer vision algorithms including affine covariant feature detectors, HOG, SIFT, MSER, kmeans, hierarchical kmeans, agglomerative information bottleneck, SLIC superpixels, and quick shift. It is written in C for efficiency and compatibility, with interfaces in MATLAB for ease of use, and detailed documentation throughout. It supports Windows, Mac OS X, and Linux. The latest version of VLFeat is 0.9.16. Changes:VLFeat 0.9.16: Added VL_COVDET() (covariant feature detectors). This function implements the following detectors: DoG, Hessian, Harris Laplace, Hessian Laplace, Multiscale Hessian, Multiscale Harris. It also implements affine adaptation, estiamtion of feature orientation, computation of descriptors on the affine patches (including raw patches), and sourcing of custom feature frame. Addet the auxiliary function VL_PLOTSS(). This is the second point update supported by the PASCAL Harvest programme. VLFeat 0.9.15: Added VL_HOG() (HOG features). Added VL_SVMPEGASOS() and a vastly improved SVM implementation. Added IHASHSUM (hashed counting). Improved INTHIST (integral histogram). Added VL_CUMMAX(). Improved the implementation of VL_ROC() and VL_PR(). Added VL_DET() (Detection Error Tradeoff (DET) curves). Improved the verbosity control to AIB. Added support for Xcode 4.3, improved support for past and future Xcode versions. Completed the migration of the old test code in toolbox/test, moving the functionality to the new unit tests toolbox/xtest. Improved credits. This is the first point update supported by the PASCAL Harvest (several more to come shortly).

About: svmstructmatlab is a MATLAB wrapper of T. Joachims' SVM^struct solver for structured output support vector machines. Changes:Adds support for Xcode 4.0 and Mac OS X 10.7 and greater.

About: Matlab code for learning probabilistic SVM in the presence of uncertain labels. Changes:Added missing dataset function (thanks to Hao Wu)

About: A Matlab implementation of Uncorrelated Multilinear Discriminant Analysis (UMLDA) for dimensionality reduction of tensor data via tensortovector projection Changes:Initial Announcement on mloss.org.
