From 0a8de79d2a858d0f72f8f29887e576306c4585ce Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Andre=CC=81s=20Masegosa=20Arredondo?= Date: Thu, 7 Jul 2016 12:05:46 +0200 Subject: [PATCH] Update on Readme.md file --- README.md | 19 +++++++++---------- 1 file changed, 9 insertions(+), 10 deletions(-) diff --git a/README.md b/README.md index 6a92d630f..f81b50026 100644 --- a/README.md +++ b/README.md @@ -22,8 +22,10 @@ AMIDST provides tailored parallel (powered by Java 8 Streams) and distributed (p #Features -* **Probabilistic Graphical Models**: Specify your model using probabilistic graphical models with latent variables -and temporal dependencies. +* **Probabilistic Graphical Models**: Specify your model using probabilistic graphical models with [latent variables](http://amidst.github.io/toolbox/examples/bnetworks.html) +and [temporal dependencies](http://amidst.github.io/toolbox/DynamicCodeExamples.html). AMIDST contains a large list of predefined latent variable models: + +![](http://amidst.github.io/toolbox/docs/web/figs/amidstModels-crop.png) * **Scalable inference**: Perform inference on your probabilistic models with powerful approximate and scalable algorithms. @@ -32,13 +34,13 @@ scalable algorithms. appropriate for learning from (massive) data streams. * **Large-scale Data**: Use your defined models to process massive data sets in a distributed -computer cluster using Apache Flink or (soon) **Apache Spark**. +computer cluster using **Apache Flink** or (soon) **Apache Spark**. * **Extensible**: Code your models or algorithms within AMiDST and expand the toolbox functionalities. Flexible toolbox for researchers performing their experimentation in machine learning. * **Interoperability**: Leverage existing functionalities and algorithms by interfacing -to other software tools such as Hugin, MOA, Weka, R, etc. +to other software tools such as [Hugin](http://amidst.github.io/toolbox/examples/bnetworks.html#sec:bns:huginlink), [MOA](http://amidst.github.io/toolbox/examples/bnetworks.html#sec:bns:moalink), Weka, R, etc. #Simple Code Example @@ -148,15 +150,12 @@ architecture and 32 cores. The size of the processed data set was measured accor

-## Distributed Scalablity using [Apache Flink](http://flink.com) +## Distributed Scalablity using Apache Flink If your data is really big and can not be stored in a single laptop, you can also learn your probabilistic model on it by using the AMIDST distributed learning engine based on -a novel and state-of-the-art [distributed message passing scheme](http://amidst.github.io/toolbox/docs/dVMP.pdf). -As detailed in this [paper](http://amidst.github.io/toolbox/docs/dVMP.pdf), we learn a probabilistic -model similar to the one detailed in the above section from 42 million samples in an Amazon's cluster with -2, 4, 8 and 16 nodes, each node containing 8 processing units. The following figure shows the scalability of -our approach under these settings. +a novel and state-of-the-art [distributed message passing scheme](http://amidst.github.io/toolbox/docs/dVMP.pdf) implemented on top +of [Apache Flink](http://flink.com). As detailed in this [paper](http://amidst.github.io/toolbox/docs/dVMP.pdf), we were able to perform inference in a billion node (i.e. 10^9) probabilistic model in an Amazon's cluster with 2, 4, 8 and 16 nodes, each node containing 8 processing units. The following figure shows the scalability of our approach under these settings.