rebordao

#!/usr/bin/env bash
:'
# VM AutoStarter

An utility to start up automatically a set of Virtual Machines 
if they are not running.

This utility was tested on a Debian 7.5 with VirtualBox 4.3.

## Usage

rebordao

"
A quicksort implementation to sort an array without using conventional functions.

It has two phases and it's a divide and conquer approach:
- partition phase
- sort phase

There is a nice video that demonstrates this concept using 
Hungarian dance at https://www.youtube.com/watch?v=ywWBy6J5gz8
"

rebordao

"
A mergesort implementation to sort an array without using conventional functions.

This algorithm has 2 stages:
- divides the array into a list of sublists where each sublist has just 1 number.
- then merge bottom-up until it outputs the sorted array.

These videos explain everything you need to know about mergesort:
- https://class.coursera.org/algo-003/lecture/1
- https://class.coursera.org/algo-003/lecture/2

rebordao

[
    {
        "Id": "28f08de53fba9115283cfcd759866c30c9ee37912a5ab968182123e4aafa3023",
        "Created": "2016-10-12T15:43:40.86284117Z",
        "Path": "/init",
        "Args": [],
        "State": {
            "Status": "running",
            "Running": true,
            "Paused": false,

rebordao

function [net hid_without_bias output] = trains_nn(net, data, targets, phonemes_id);
% This function trains a network by backpropagation, by backpropagation
% through time or by connectionist temporal classification.
%
% INPUTS:
% net contains the topology and the user-defined training parameters;
% data is the train data
% targets are the real targets
% phonemes_id is the indices of the phonemes in the dictionary file
%

rebordao

function batch_nn(architecture, nr_hidden_units, learning_rate, mom, regularization_term, verbose)
% This function implements a speech recognizer for TIMIT samples.
% Architecture: a FFNN trained by Back Propagation; a RNN trained by BPTT or CTC; or a Reservoir trained via ctc.
% 
% INPUTS:
% architecture can be ffnn, rnn or reservoir
% nr_hidden_units is the nr of hidden units
% learning_rate is the learning rate for the weights' update
% mom is the momemtum rate
% regularization_term defines the regularization