Doubt, for narrator and orchestra

Doubt is a single-movement composition of roughly twelve minutes for narrator and orchestra (woodwinds, horns, and trumpets in pairs, timpani, percussion, strings). The piece explores the controversial issue of capital punishment. The text was compiled from resources found on the websites of Death Penalty Information Center (http://www.deathpenaltyinfo.org) and Anti-Death Penalty Information (http://www.antideathpenalty.org), as well as excerpts from the Bible. Doubt was conceived of as a dramatic work in which a narrator recites factual information in a direct and unemotional manner and the orchestra provides a response to the mixed emotions elicited by the text. The list of dates and case summaries presented in the middle section of the piece seemed most powerful and effective when recited in a natural speaking voice, which is why I chose not to set the text as song. Also, I chose the orchestral medium rather than a chamber setting because the nature of the topic demanded a larger range of colors and combinations, as well as a louder, fuller sound. Much of the music was composed while deciding which texts to include. Thus the music influenced the choice of text as much as the text suggested the musical setting. The four formal divisions of the piece are delineated primarily by the text. The first section is an orchestral introduction representing various emotional perspectives suggested by the texts. The narrator begins the second section with a Biblical verse over sparse orchestration. The third and main section of the piece begins with a new melody in the low strings that is closely related to the harmonic organization of the piece. The narrator lists dates of convictions, executions, exonerations and facts related to doubtful cases. The third section and the narration conclude with another brief passage from the Bible. The fourth section is a dramatic orchestral coda, bringing back the opening harmonies of juxtaposed perfect fifths. The final chord is full of tension and discord, reflecting the oppositions inherent in the topic of capital punishment: life vs. death, sympathy vs. reproach, pain vs. hope, but above all, doubt about guilt vs. innocence.

Learning Binary Code Representations for Effective and Efficient Image Retrieval

The size of online image datasets is constantly increasing. Considering an image dataset with millions of images, image retrieval becomes a seemingly intractable problem for exhaustive similarity search algorithms. Hashing methods, which encodes high-dimensional descriptors into compact binary strings, have become very popular because of their high efficiency in search and storage capacity. In the first part, we propose a multimodal retrieval method based on latent feature models. The procedure consists of a nonparametric Bayesian framework for learning underlying semantically meaningful abstract features in a multimodal dataset, a probabilistic retrieval model that allows cross-modal queries and an extension model for relevance feedback. In the second part, we focus on supervised hashing with kernels. We describe a flexible hashing procedure that treats binary codes and pairwise semantic similarity as latent and observed variables, respectively, in a probabilistic model based on Gaussian processes for binary classification. We present a scalable inference algorithm with the sparse pseudo-input Gaussian process (SPGP) model and distributed computing. In the last part, we define an incremental hashing strategy for dynamic databases where new images are added to the databases frequently. The method is based on a two-stage classification framework using binary and multi-class SVMs. The proposed method also enforces balance in binary codes by an imbalance penalty to obtain higher quality binary codes. We learn hash functions by an efficient algorithm where the NP-hard problem of finding optimal binary codes is solved via cyclic coordinate descent and SVMs are trained in a parallelized incremental manner. For modifications like adding images from an unseen class, we propose an incremental procedure for effective and efficient updates to the previous hash functions. Experiments on three large-scale image datasets demonstrate that the incremental strategy is capable of efficiently updating hash functions to the same retrieval performance as hashing from scratch.