Posts

The second week of the Google Summer of Code project with ReservoirComputing.jl dealt with the implementation of a Support Vector based regression for the Echo State Network model, resulting in the Support Vector Echo-State Machine (SVESM). In this post we will quickly touch on the theory behind Support Vector Regression (SVR), and then we will see the results of the implementation into the library. At the end a couple of examples are given, togheter with a comparison between SVR and SVESMs....

The first week tackled the implementation of different kind of linear regression for the creation of the last layer in the Echo State Network. More specifically were added the possibility to add a \( l_1 \) regularization to the loss function (Lasso regression), both \( l_1 \) and \( l_2 \) regularizations (Elastic Net regression) and also added the possibility to choose the Huber loss function. As in the last post we will start from a brief theoretical background to explain the code and then we will showcase some examples taken from the literature....

This post is meant to work as an high level introduction to the concept of Reservoir Computing, using the Julia package ReservoirComputing.jl as example tool. This package is a work in progress and it is currently the main project I am working on as part of the Google Summer of Code program. Future posts are going to further explain the various implementations and improvements to the code by means of comparisons with the literature and examples....