Rafael Valle

Missing Data Imputation for Supervised Learning

ABSTRACT Missing data imputation can help improve the performance of prediction models in situations where missing data hide useful information. This paper compares methods for imputing missing categorical data for supervised classification tasks. We experiment on two machine learning benchmark datasets with missing categorical data, comparing classifiers trained on non-imputed (i.e., one-hot encoded) or imputed data with different levels of additional missing-data perturbation. We show imputation methods can increase predictive accuracy in the presence of missing-data perturbation, which can actually improve prediction accuracy by regularizing the classifier. We achieve results comparable to the state-of-the-art on the Adult dataset with missing-data perturbation and -nearest-neighbors (-NN) imputation.

ABROA: Audio-based room-occupancy analysis using Gaussian mixtures and Hidden Markov models

This paper outlines preliminary steps towards the development of an audio-based room-occupancy analysis model. Our approach borrows from speech recognition tradition and is based on Gaussian Mixtures and Hidden Markov Models. We analyse possible challenges encountered in the development of such a model, and offer several solutions including feature design and prediction strategies. We provide results obtained from experiments with audio data from a retail store in Palo Alto, California. Model assessment is done via leave-two-out Bootstrap and model convergence achieves good accuracy, thus representing a contribution to multimodal people counting algorithms.

Specification Mining for Machine Improvisation with Formal Specifications

We address the problem of mining musical specifications from a training set of songs and using these specifications in a machine improvisation system capable of generating improvisations imitating a given style of music. Our inspiration comes from control improvisation, which combines learning and synthesis from formal specifications. We mine specifications from symbolic musical data with musical and general usage patterns. We use the mined specifications to ensure that an improvised musical sequence satisfies desirable properties given a harmonic context and phrase structure. We present a specification mining strategy based on pattern graphs and apply it to the problem of supervising the improvisation of blues songs. We present an analysis of the mined specifications and compare the results of improvisations generated with and without specifications.

Symbolic Music Similarity Using Neuronal Periodicity and Dynamic Programming

We introduce NP-MUS, a symbolic music similarity algorithm tailored for polyphonic music with continuous representations of pitch and duration. The algorithm uses dynamic programming and a cost function that relies on a mathematical model of tonal fusion based on neuronal periodicity detection mechanisms. This paper reviews the general requirements of melodic similarity and offers a similarity method that better addresses contemporary and non-traditional music. We provide experiments based on monophonic and polyphonic excerpts inspired by spectral music and Iannis Xenakis.

Control Improvisation with Probabilistic Temporal Specifications

We consider the problem of generating randomized control sequences for complex networked systems typically actuated by human agents. Our approach leverages a concept known as control improvisation, which is based on a combination of data-driven learning and controller synthesis from formal specifications. We learn from existing data a generative model (for instance, an explicit-duration hidden Markov model, or EDHMM) and then supervise this model in order to guarantee that the generated sequences satisfy some desirable specifications given in Probabilistic Computation Tree Logic (PCTL). We present an implementation of our approach and apply it to the problem of mimicking the use of lighting appliances in a residential unit, with potential applications to home security and resource management. We present experimental results showing that our approach produces realistic control sequences, similar to recorded data based on human actuation, while satisfying suitable formal requirements.