Publicado em: 10/05/2016

Prof. Nuno Marques da Univ. Nova de Lisboa no INF

Curso “Processing Text and Financial Data for Deep Learning Methods” no INF

De 9 a 13 de maio, o Prof. Nuno Marques da Universidade Nova de Lisboa estará ministrando um curso intensivo teórico e prático de 15 horas sobre Deep Learning, que acontecerá à tarde no Instituto de Informática.

Informações sobre o Curso:

Título: “Processing Text and Financial Data for Deep Learning Methods”

Financial data provides a valuable up-to-date knowledge of the world economy. However, it is presented in extremely large data volumes, in diverse formats and is constantly being updated at a high speed.

This short course overviews recent developments in neural networks and how they can be used to help text and data mining tasks over financial text & data sources. We start by introducing standard backprop neural networks. Then the overfitting problem is used to clarify the advantages of recent deep-learning methods. Text streams are ilustrated with the deep-learning model Word2Vec (Mikolov et al, 2013) for the continuous bag-of-word model (CBOW) and skip-ngram. Then non-supervised neural networks are ilustrated with UbiSOM, a novel variant of the Self-Organized Map (SOM) that was tailored for streaming environments. This approach allows ambient intelligence solutions using multidimensional clustering over a continuous data stream to provide continuous exploratory data analysis. The average quantization error and other neuron utility over time are used to allow the model to retain an indefinite plasticity and to cope with changes within a multidimensional data stream.

Theoretical presentation will be integrated with pratical sessions.

Both octave and unix text mining tools (such as ‘tr’, ‘gawk’ and ‘sed’) will be used for preparing and testing several examples with word2vec and UbiSOM.

Recent results and open issues regarding information retrieval from available data in financial data or other data streams will be discussed.