Entradas por fecha "Junio 2017"

Second public interface in digitalocean

In APSL we use the cloud provider digitalocean (DO onwards) for some support services and we are really happy with it. It’s not at the same level in services of AWS but allows us reduce costs in small self-managed services at the cloud.

One limitation with a "droplet" (this is how call to an instance in DO) is the few options for the configuration. DO assigns you a fixed IPv4 in the host and it’s now allowed add extra network interfaces. This is a limitation when you want to do something that is outside the norm, for example want to serve two https sites without SNI, or the problem that APSL found this week, the Gitlab Pages service needs a second public interface to configure a secondary webserver.

Leer mas

Using Principal Component Analysis (PCA) for data Explorer. Step by Step

In this contribution we show, through some toy examples, how to use Matrix Factorization techniques to analyze multivariate data sets in order to obtain some conclusions from them that may help us to take decisions. In particular, we explain how to employ the technique of Principal Component Analysis (PCA) to reduce the dimensionality of the space of variables.

Leer mas

Análisis de Series Temporales Usando Redes Neuronales Recurrentes.

En éste artículo se introduce el uso de Redes Neuronales Recurrentes (RNN) como nuevo enfoque para tratar el problema del análisis y predicción de series temporales. Como casos de estudio y para demostrar el grado de éxito del uso de RNN en éste contexto, aplicamos este enfoque al estudio del consumo eléctrico en la población de Sóller (Mallorca) y en el estudio del consumo eléctrico en la isla de Tenerife. El objetivo es mostrar cómo, con esta metodología se puede predecir el consumo eléctrico de las poblaciones con un grado de precisión que ronda el 93%.Para ello, proponemos el uso de un tipo de red neuronal recurrente, conocida como “Long Short Term Memory Network” (LSTM).

Leer mas