We present a new technique called “t-SNE” that visualizes high-dimensional data by giving each datapoint a location in a two or three-dimensional map. The technique is a variation of Stochastic Neighbor Embedding (Hinton and Roweis, 2002) that is much easier to optimize, and produces significantly better visualizations by reducing the tendency to crowd points together in the center of the map. t-SNE is better than existing techniques at creating a single map that reveals structure at many different scales. This is particularly important for high-dimensional data that lie on several different, but related, low-dimensional manifolds, such as images of objects from multiple classes seen from multiple viewpoints. For visualizing the structure of very large datasets, we show how t-SNE can use random walks on neighborhood graphs to allow the implicit structure of all of the data to influence the way in which a subset of the data is displayed. We illustrate the performance of t-SNE on a wide variety of datasets and compare it with many other non-parametric visualization techniques, including Sammon mapping, Isomap, and Locally Linear Embedding. The visualizations produced by t-SNE are significantly better than those produced by the other techniques on almost all of the datasets.

/pdf/visualizing-data-using-t-sne-2w1n304iq0.pdf

Visualizing Data using t-SNE

Three parallel algorithms for classical molecular dynamics are presented. The first assigns each processor a fixed subset of atoms; the second assigns each a fixed subset of inter-atomic forces to compute; the third assigns each a fixed spatial region. The algorithms are suitable for molecular dynamics models which can be difficult to parallelize efficiently—those with short-range forces where the neighbors of each atom change rapidly. They can be implemented on any distributed-memory parallel machine which allows for message-passing of data between independently executing processors. The algorithms are tested on a standard Lennard-Jones benchmark problem for system sizes ranging from 500 to 100,000,000 atoms on several parallel supercomputers--the nCUBE 2, Intel iPSC/860 and Paragon, and Cray T3D. Comparing the results to the fastest reported vectorized Cray Y-MP and C90 algorithm shows that the current generation of parallel machines is competitive with conventional vector supercomputers even for small problems. For large problems, the spatial algorithm achieves parallel efficiencies of 90% and a 1840-node Intel Paragon performs up to 165 faster than a single Cray C9O processor. Trade-offs between the three algorithms and guidelines for adapting them to more complex molecular dynamics simulations are also discussed.

Fast parallel algorithms for short-range molecular dynamics

https://axon.cs.byu.edu/~martinez/classes/678/Presentations/Yao.pdf

Extreme learning machine: Theory and applications

Probability Distributions.- Linear Models for Regression.- Linear Models for Classification.- Neural Networks.- Kernel Methods.- Sparse Kernel Machines.- Graphical Models.- Mixture Models and EM.- Approximate Inference.- Sampling Methods.- Continuous Latent Variables.- Sequential Data.- Combining Models.

Pattern Recognition and Machine Learning

Data Mining - Concepts and Techniques.

We propose a method of function approximation by radial basis function networks. We will demonstrate that this approximation method can be improved by a pre-treatment of data based on a linear model. This approximation method will be applied to option pricing. This choice justifies itself through the known nonlinear nature of the behaviour of options price and through the effective contribution of the pre-treatment proposed for the implementation of radial basis function networks in this field.

Approximation by radial basis function networks

The quantitative steganalysis problem aims at estimating the amount of payload embedded inside a document. In this paper, JPEG images are considered, and by the use of a re-embedding based methodology, it is possible to estimate the number of original embedding changes performed on the image by a stego source and to slightly improve the estimation regarding classical quantitative steganalysis methods. The major advance of this methodology is that it also enables to obtain a confidence interval on this estimated payload. This confidence interval then permits to evaluate the difficulty of an image, in terms of steganalysis by estimating the reliability of the output. The regression technique comes from the OP-ELM and the reliability is estimated using linear approximation. The methodology is applied with publicly available stego algorithms, regression model and large databases of images. The methodology is generic and can be used for any quantitative steganalysis problem of this class.

Using multiple re-embeddings for quantitative steganalysis and image reliability estimation

This paper proposes a block solution method for the Extreme Learning Machine. It combines the speed of a direct non-iterative solver with minimal memory requirements. The method is suitable for edge computing scenarios running on a mobile device with GPU acceleration. The implementation tested on the GPU of iPad Pro outperforms a laptop CPU, and trains a 19,000-neuron model using under one gigabyte of memory. It confirms the feasibility of Big Data analysis on modern mobile devices.

High-Performance ELM for Memory Constrained Edge Computing Devices with Metal Performance Shaders

Nowadays, ocean observatory networks, which gather and provide multidisciplinary, long-term, 3D continuous marine observations at multiple temporal spatial scales, play a more and more important role in ocean investigations. In this paper, we try to develop a portable smart device with online fish detection and tracking strategies by ARM7 microprocessor for ocean observatory networks, combining the deformable fish body description with compressive sensing together. A fish detection system is introduced to represent highly variable objects using mixtures of multiscale deformable fish body description which include three parts:(1) root filter, which describes the global contour feature of underwater fish.(2) fish body part filter, which capture finer resolution features of underwater fish. (3) spatial model, which describes the relationship between the fish body part filters and the root filter. And then, we apply a simple yet effective, real-time and robust of fast compressive tracking algorithm with an appearance model to focus on the motion trajectories of underwater fish. It has been shown in the simulation experiments that the developed scheme of this paper achieves consistent performance improvements on online fish tracking for ocean observatory network.

Amaury Lendasse

Papers

Approximation by radial basis function networks

Using multiple re-embeddings for quantitative steganalysis and image reliability estimation

High-Performance ELM for Memory Constrained Edge Computing Devices with Metal Performance Shaders

Online fish tracking with portable smart device for ocean observatory network

Advances in extreme learning machines (ELM2015)