Query-Sensitive Embeddings

A common problem in many types of databases is retrieving the most similar matches to a query object. Finding those matches in a large database can be too slow to be practical, especially in domains where objects are compared using computationally expensive similarity (or distance) measures. This paper proposes a novel method for approximate nearest neighbor retrieval in such spaces. Our method is embedding-based, meaning that it constructs a function that maps objects into a real vector space. The mapping preserves a large amount of the proximity structure of the original space, and it can be used to rapidly obtain a short list of likely matches to the query. The main novelty of our method is that it constructs, together with the embedding, a query-sensitive distance measure that should be used when measuring distances in the vector space. The term "query-sensitive" means that the distance measure changes depending on the current query object. We report experiments with an image database of handwritten digits, and a time-series database. In both cases, the proposed method outperforms existing state-of-the-art embedding methods, meaning that it provides significantly better trade-offs between efficiency and retrieval accuracy.

Dense spectrum of resonances and particle capture in a near-black-hole metric

We show that a quantum scalar particle in the gravitational field of a massive body of radius R which slightly exceeds the Schwarzschild radius rs, possesses a dense spectrum of narrow resonances. Their lifetimes and density tend to infinity in the limit R?rs. We determine the cross section of the particle capture into these resonances and show that it is equal to the absorption cross section for a Schwarzschild black hole. Thus, a nonsingular static metric acquires black-hole properties before the actual formation of a black hole.

Impact of Metric Selection on Wireless DeAuthentication DoS Attack Performance

DeAuthentication Denial of Service attacks in Public Access WiFi operate by exploiting the lack of authentication of management frames in the 802.11 protocol. Detection of these attacks rely almost exclusively on the selection of appropriate thresholds. In this work the authors demonstrate that there are additional, previously unconsidered, metrics which also influence DoS detection performance. A method of systematically tuning these metrics to optimal values is proposed which ensures that parameter choices are repeatable and verifiable.

Development and assessment of a robust optimization metric for WAMPAC modeling

A PMU based WAMS is to be placed on a weakly coupled section of distribution grid, with high levels of distributed generation. In anticipation of PMU data a Siemens PSS/E model of the electrical environment has been used to return similar data to that expected from the WAMS. This data is then used to create a metric that reflects optimization, control and protection in the region. System states are iterated through with the most desirable one returning the lowest optimization metric, this state is assessed against the one returned by PSS/E under normal circumstances. This paper investigates the circumstances that trigger SPS in the region, through varying generation between 0 and 110% and compromising the network through line loss under summer minimum and winter maximum conditions. It is found that the optimized state can generally tolerate an additional 2 MW of generation (3% of total) before encroaching the same thresholds and in one instance moves the triggering to 100% of generation output.

Dense spectrum of resonances and spin-1/2 particle capture in a near-black-hole metric

We show that a spin-1/2 particle in the gravitational field of a massive body of radius R which slightly exceeds the Schwarzschild radius rs, possesses a dense spectrum of narrow resonances. Their lifetimes and density tend to infinity in the limit R → rs. We determine the cross section of the particle capture into these resonances and show that it is equal to the spin-1/2 absorption cross section for a Schwarzschild black hole. Thus black-hole properties may emerge in a non-singular static metric prior to the formation of a black hole.

Efficient RkNN Retrieval with Arbitrary Non-Metric Similarity Measures

A RkNN query returns all objects whose nearest k neighbors
contain the query object. In this paper, we consider RkNN
query processing in the case where the distances between
attribute values are not necessarily metric. Dissimilarities
between objects could then be a monotonic aggregate of dissimilarities
between their values, such aggregation functions
being specified at query time. We outline real world cases
that motivate RkNN processing in such scenarios. We consider
the AL-Tree index and its applicability in RkNN query
processing. We develop an approach that exploits the group
level reasoning enabled by the AL-Tree in RkNN processing.
We evaluate our approach against a Naive approach
that performs sequential scans on contiguous data and an
improved block-based approach that we provide. We use
real-world datasets and synthetic data with varying characteristics
for our experiments. This extensive empirical
evaluation shows that our approach is better than existing
methods in terms of computational and disk access costs,
leading to significantly better response times.

Bounds on edit metric codes with combinatorial DNA constraints

The design of a large and reliable DNA codeword library is a key problem in DNA based computing. DNA codes, namely sets of fixed length edit metric codewords over the alphabet {A, C, G, T}, satisfy certain combinatorial constraints with respect to biological and chemical restrictions of DNA strands. The primary constraints that we consider are the reverse--complement constraint and the fixed GC--content constraint, as well as the basic edit distance constraint between codewords. We focus on exploring the theory underlying DNA codes and discuss several approaches to searching for optimal DNA codes. We use Conway's lexicode algorithm and an exhaustive search algorithm to produce provably optimal DNA codes for codes with small parameter values. And a genetic algorithm is proposed to search for some sub--optimal DNA codes with relatively large parameter values, where we can consider their sizes as reasonable lower bounds of DNA codes. Furthermore, we provide tables of bounds on sizes of DNA codes with length from 1 to 9 and minimum distance from 1 to 9.

From Word Embeddings to Large Vocabulary Neural Machine Translation

Dans ce mémoire, nous examinons certaines propriétés des représentations distribuées de mots et nous proposons une technique pour élargir le vocabulaire des systèmes de traduction automatique neurale. En premier lieu, nous considérons un problème de résolution d'analogies bien connu et examinons l'effet de poids adaptés à la position, le choix de la fonction de combinaison et l'impact de l'apprentissage supervisé. Nous enchaînons en montrant que des représentations distribuées simples basées sur la traduction peuvent atteindre ou dépasser l'état de l'art sur le test de détection de synonymes TOEFL et sur le récent étalon-or SimLex-999. Finalament, motivé par d'impressionnants résultats obtenus avec des représentations distribuées issues de systèmes de traduction neurale à petit vocabulaire (30 000 mots), nous présentons une approche compatible à l'utilisation de cartes graphiques pour augmenter la taille du vocabulaire par plus d'un ordre de magnitude. Bien qu'originalement développée seulement pour obtenir les représentations distribuées, nous montrons que cette technique fonctionne plutôt bien sur des tâches de traduction, en particulier de l'anglais vers le français (WMT'14).