Nuno Marques

Universidade Nova de Lisboa, Computer Science, Faculty Member

Followers

Following

Public Views

Interests

Uploads

Papers by Nuno Marques

The ubiquitous self-organizing map for non-stationary data streams

Journal of Big Data, Dec 1, 2015

The Internet of things promises a continuous flow of data where traditional database and data-min... more The Internet of things promises a continuous flow of data where traditional database and data-mining methods cannot be applied. This paper presents improvements on the Ubiquitous Self-Organized Map (UbiSOM), a novel variant of the well-known Self-Organized Map (SOM), 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 average neuron utility over time are proposed and used to estimating the learning parameters, allowing the model to retain an indefinite plasticity and to cope with changes within a multidimensional data stream. We perform parameter sensitivity analysis and our experiments show that UbiSOM outperforms existing proposals in continuously modeling possibly non-stationary data streams, converging faster to stable models when the underlying distribution is stationary and reacting accordingly to the nature of the change in continuous real world data streams.

format_quoteUbiSOM outperforms other online SOM methods in modeling non-stationary data streams and converges faster to stable models.format_quote

Download

Stock Market Series Analysis Using Self-Organizing Maps

In this work a new clustering technique is implemented and tested. The proposed approach is based... more In this work a new clustering technique is implemented and tested. The proposed approach is based on the application of a SOM (self-organizing map) neural network and provides means to cluster U-MAT aggregated data. It relies on a flooding algorithm operating on the U-MAT and resorts to the Calinski and Harabask index to assess the depth of flooding, providing an adequate number of clusters. The method is tuned for the analysis of stock market series. Results obtained are promising although limited in scope.

Download

Data Streams for Unsupervised Analysis of Company Data

Progress in Artificial Intelligence, 2021

Finding Compositional Rules for Determining the Semantic Orientation of Phrases

Lecture Notes in Computer Science, 2016

The semantic compositionality principle states that the meaning of an expression can be determine... more The semantic compositionality principle states that the meaning of an expression can be determined by its parts and the way they are put together. Based on that principle, this paper presents a method for finding the set of compositional rules that best explain the positive, negative, and neutral semantic orientation (SO) of two-word phrases, in terms of the SO of its words. For instance, the phrase “fake contract” has a negative SO. A corpus was built for evaluating the proposed method and several experiences are reported. We also use the conditional probability as a reliability measure of the compositional rules. The reliability of the learned rules ranges from 60.44 % for verb-noun phrases to 100 % for adjective-adjective phrases.

Recognition of Good, Bad, and Neutral News Headlines in Portuguese

This paper investigates the classification of news headlines as positive, negative, and neutral. ... more This paper investigates the classification of news headlines as positive, negative, and neutral. A news headline is positive if is associated with good things, negative if it is associated with bad things, and neutral in the remaining cases. The classification of a news headline is predicted using a supervised approach. The experiments show an accuracy that ranges from 59.00% to 63.50% when argument1-verb-argument2 relations are combined with other features. The accuracy ranges from 57.50% to 62.5% when these relations are not used.

Download

Applying a Part-of-Speech Tagger to Postal Address Detection on the Web

In this paper we show how a POS-tagger can be successfully adapted to a real world information re... more In this paper we show how a POS-tagger can be successfully adapted to a real world information retrieval system capable of extracting postal addresses from the Internet. We develop a particular tag-set for this system. Then we present and discuss the results acquired with the developed postal address tag-set. We conclude the paper by presenting a short description of the final IR system for Web postal address retrieval and drawing some conclusions.

Download

An extension of the core method for continuous values: Learning with probabilities

This paper proposes an extension to the neuro-symbolic core method useful when observations are e... more This paper proposes an extension to the neuro-symbolic core method useful when observations are expressed by continuous values. Some theoretical results are presented regarding the learning process over these observations. An illustrative example is reported, demonstrating the problems of the original approach and justifying how this extension can overcome them. Results of the extended approach on irregular con-tinuous values (simulating probabilistic data) show similar results to the original core method on clean symbolic data and point to the validity of the approach.

Download

Agent's Belief: A Stochastic Approach

In this paper we present a theoretical model for evaluating agent's belief. We model the agen... more In this paper we present a theoretical model for evaluating agent's belief. We model the agent's belief as a threshold function of agent's certainty. The main problem is that agent's certainty is unobservable. This model is based on results from different tests. A propagation algorithm is presented. We also discuss an example for determining a statistical agent's belief in a probability model.

Towards Encoding Background Knowledge with Temporal Extent into Neural Networks

Lecture Notes in Computer Science, 2010

Neuro-symbolic integration merges background knowledge and neural networks to provide a more effe... more Neuro-symbolic integration merges background knowledge and neural networks to provide a more effective learning system. It uses the Core Method as a means to encode rules. However, this method has several drawbacks in dealing with rules that have temporal extent. First, it demands some interface with the world which buffers the input patterns so they can be represented all at once. This imposes a rigid limit on the duration of patterns and further suggests that all input vectors be the same length. These are troublesome in domains where one would like comparable representations for patterns that are of variable length (e.g. language). Second, it does not allow dynamic insertion of rules conveniently. Finally and also most seriously, it cannot encode rules having preconditions satisfied at non-deterministic time points-an important class of rules. This paper presents novel methods for encoding such rules, thereby improves and extends the power of the state-of-the-art neuro-symbolic integration.

Download

Ubiquitous Self-Organizing Map: Learning Concept-Drifting Data Streams

Advances in Intelligent Systems and Computing, 2015

The Internet of Things promises a continuous flow of data where traditional database and data-min... more The Internet of Things promises a continuous flow of data where traditional database and data-mining methods cannot be applied. This paper presents a novel variant of the well-known Self-Organized Map (SOM), called Ubiquitous SOM (UbiSOM), that is being 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 over time is used for estimating the learning parameters, allowing the model to retain an indefinite plasticity and to cope with concept drift within a multidimensional stream.

FeaSANNT - an Embedded Evolutionary Feature Selection Approach for Neural Network Classifiers

This paper presents FeaSANNT, an evolutionary feature selec-tion and weight training pro-cedure f... more This paper presents FeaSANNT, an evolutionary feature selec-tion and weight training pro-cedure for neural network clas-sifi ers. FeaSANNT exploits the global nature of the evolution-ary search to avoid sub-optimal peaks of performance. The nov-elty of the method lies in the implementation of the embed-ded approach in an evolution-ary feature selection paradigm. Such choice minimises the com-putational overheads and eases the algorithm design. FeaSAN-NT is used to train a multi-layer perceptron classifi er on six real-world benchmark problems. FeaSANNT achieves accurate and robust learning results with sig-nifi cant reduction of the feature set. Experimental comparison shows FeaSANNT outperforms three standard state-of-the-art feature selection methods.

Automatic Detection of Meddies Through Texture Analysis of Sea Surface Temperature Maps

Lecture Notes in Computer Science, 2005

Abstract. A new machine learning approach is presented for automatic detec-tion of Mediterranean ... more

Uma Metodologia para a Modelac? ?ao Estat'istica da Subcategorizac? ?ao Verbal

Border Detection on Remote Sensing Satellite Data Using Self-Organizing Maps

Lecture Notes in Computer Science, 2003

Page 1. Border Detection on Remote Sensing Satellite Data Using Self-Organizing Maps Nuno C. Marq... more

An Extension of Self-organizing Maps to Categorical Data

Progress in Artificial Intelligence, 2005

Self-organizing maps (SOM) have been recognized as a powerful tool in data exploratoration, espec... more Self-organizing maps (SOM) have been recognized as a powerful tool in data exploratoration, especially for the tasks of clustering on high dimensional data. However, clustering on categorical data is still a challenge for SOM. This paper aims to extend standard SOM to handle feature values of categorical type. A batch SOM algorithm (NCSOM) is presented concerning the dissimilarity measure and update method of map evolution for both numeric and categorical features simultaneously.

Download

Simulating Price Interactions by Mining Multivariate Financial Time Series

This position paper proposes a framework based on a feature clustering method using Emergent Self... more This position paper proposes a framework based on a feature clustering method using Emergent Self-Organizing Maps over streaming data (Ubi-SOM) and Ramex-Forum-a sequence pattern mining model for financial time series modeling based on observed instantaneous and long term relations over market data. The proposed framework aims at producing realistic monte-carlo based simulations of an entire portfolio behavior over distinct market scenarios, obtained from models generated by these two approaches.

Download

Sequential pattern mining of price interactions

The computational analysis of large quantities of data is an important asset for the economic stu... more The computational analysis of large quantities of data is an important asset for the economic study of interactions among social agents. However, most of available frequent pattern discovery techniques result in a huge number of rules and scalability problems that end up requiring unnecessary subjectivity in data interpretation. This work presents Ramex-Forum, a visualization technique that can highlight important relations often hidden in economic data. A case study using recent asset prices on global economic data confirm the usefulness of the approach for expressingeconomic influence cues as poly-trees.

Download

The Role of Language Registers in Polarity Propagation

Lecture Notes in Computer Science, 2012

This paper presents work on the automatic creation of a polarity lexicon based on a lexical-seman... more This paper presents work on the automatic creation of a polarity lexicon based on a lexical-semantic network. During this work, we noticed that the language registers of a relation should be considered in polarity propagation. After analysing the possible registers and performing some experiments, our intuition was confirmed-there are registers that should invert the transmitted polarity (irony), while others always transmit negative polarity (pejorative, disparaging).

Download

Utilização da programação declarativa para processamento do CETEMPúblico

… sobre a Linguateca, Actas do encontro …

... Anabela Barreiro, Belinda Maia, Cristina Mota, Débora Oliveira, Eckhard Bick, Elisabete Ranch... more

Polaris: POrtuguese lexicon acquisition and retrieval interactive system

The Practical Applications of …, 1994

The present paper is intended for presenting a Portuguese lexicon acquisition and retrieval syste... more

Nuno Marques

Uploads

Papers by Nuno Marques

Log In