1,721,051 research outputs found
Locality sensitive hashing schemes, similarities, and distortion (invited talk)
No full textLocality sensitive hashing (LSH) is a key algorithmic tool that lies at the heart of many information retrieval and machine learning systems [1, 2, 8]. LSH schemes are used to sketch large objects (e.g., Web pages, fields of flowers, or - more generally - sets and vectors) into fingerprints of few bits each; the fingerprints are then used to quickly, and approximately, reconstruct some similarity relation between the objects. A LSH scheme for a similarity (or, analogously, for a distance) can significantly improve the computational cost of many algorithmic primitives (e.g., nearest neighbor search, and clustering). For this reason, in the last two decades, researchers have tried to understand which similarities admit efficient LSH schemes: such schemes were obtained for many similarities [1-3, 7-9], while the non-existence of LSH schemes was proved for a number of other similarities [3]. In our talk, we will introduce the class of LSH-preserving transformations [4] (functions that, when applied to a similarity that admits a LSH scheme, return a similarity that also admits such a scheme). We will give a characterization of this class of functions: they are precisely the probability generating functions, up to scaling. We will then show how this characterization can be used to construct LSH schemes for a number of well-known similarities. We will then discuss a notion of similarity distortion [6], in order to deal with similarities which are known to not admit LSH schemes - this notion aims to determine the minimum distortions that these similarities have to be subject of, before starting to admit a LSH scheme. We will introduce a number of general theoretical tools that can be used to determine the optimal distortions of some important classes of similarities. Finally, we will consider the computational problem of checking whether a similarity admits a LSH scheme [5], showing that, unfortunately, this problem is computationally hard in a very strong sense
18F-FDG-PET/CT.
No full textIn the assessment of brain tumors, in spite of its low sensitivity in some histological exams, chiefly low-grade lesions, its prognostic value remains of high clinical impact. Moreover, the reliability of [18F]FDG in examining not only the tumor itself, but also the functional state of the whole brain, makes this tracer a valuable tool for treatment decisions and patient management, even nowadays when new tracers (especially amino-acids) are available. In addition, [18F]FDG has a role in the differential diagnosis between relapse and necrosis when assessing aggressive tumors and to establish dedifferentiation in low-grade lesions. With the growing of available therapies, another emerging application of [18F]FDG is the monitoring of response to treatment, even though more evidence is needed to assess the best scanning time
Asymptotic Behavior of Sequence Models
No full textIn this paper we study the limiting dynamics of a sequential process that generalizes Pólya's urn. This process has been studied also in the context of language generation, discrete choice, repeat consumption, and models for the web graph. The process we study generates future items by copying from past items. It is parameterized by a sequence of weights describing how much to prefer copying from recent versus more distant locations. We show that, if the weight sequence follows a power law with exponent α ĝ [0, 1), then the sequences generated by the model tend toward a limiting behavior in which the eventual frequency of each token in the alphabet attains a limit. Moreover, in the case α > 2, we show that the sequence converges to a token being chosen infinitely often, and each other token being chosen only constantly many times
The Gibbs-Rand Model
No full textDue to its many applications, the clustering ensemble problem has been subject of intense algorithmic study over the last two decades. The input to this problem is a set of clusterings; its goal is to output a clustering that minimizes the average distance to the input clusterings. In this paper, we propose, to the best of our knowledge, the first generative model for this problem. Our Gibbs-like model is parameterized by a center clustering, and by a scale; the probability of a particular clustering decays exponentially with its scaled Rand distance to the center clustering.For our new model, we give polynomial-time algorithms forcenter dot sampling, when the center clustering has a constant number of clusters andcenter dot reconstruction, when the scale parameter is small.En route, we establish several interesting properties of our model. Our work shows that the combinatorial structure of a Gibbs-like model for clusterings is more intricate and challenging than the corresponding and well-studied (Mallows) model for permutations
Comparison of International Consensus Guidelines versus 18-FDG PET in detecting malignancy of intraductal papillary mucinous neoplasms of the pancreas.
No full textOBJECTIVE:
To assess the reliability of the International Consensus Guidelines (ICG) and 18-fluorodeoxyglucose positron emission tomography (PET) in distinguishing benign from malignant intraductal papillary mucinous neoplasms (IPMNs) of the pancreas.
BACKGROUND:
Since 2006 the ICG have been used to choose immediate surgery or surveillance for IPMN patients, but their low specificity increases the number of benign IPMNs that undergo resective surgery. PET has proved highly sensitive and specific in detecting malignancy in cystic neoplasms of the pancreas, including IPMNs.
METHODS:
Patients suspected with IPMNs of the pancreas seen at our Department from January 1989 to July 2010 were identified and classified as cases of main duct, mixed type and branch type IPMN. The indication for resection or surveillance was verified a posteriori for all patients according to the ICG. PET was considered positive for a Standardized Uptake Value ≥2.5. Surveillance included clinical examination, laboratory tests, CA 19-9 serum levels, and computed tomography and/or magnetic resonance and magnetic resonance cholangiopancreatography every 6 months for 2 years and yearly thereafter. Endoscopic ultrasound was rarely performed. PET was repeated in clinically or radiologically suspect cases, or if tumor markers increased.
RESULTS:
Sixty-one main duct or mixed type and 101-branch type IPMNs were included in the study. A histological diagnosis was available for 81 of 162 patients, missing for 1 locally advanced IPMN, whereas 62 patients are under surveillance and it proved impossible to contact 18. Conservative surgery was performed in 16 of 68 patients with benign IPMNs. The sensitivity, specificity, positive and negative predictive value, and accuracy of the ICG in detecting malignancy were 93.2, 22.2, 59.4, 72.7, and 61.2, whereas for PET they were 83.3, 100, 100, 84.6, and 91.3.
CONCLUSIONS:
PET is more accurate than the ICG in distinguishing benign from malignant (invasive and noninvasive) IPMNs. Prophylactic IPMN resection in young patients fit for surgery should be guided by the ICG, whereas PET should be performed in older patients, cases at increased surgical risk, or when the feasibility of parenchyma-sparing surgery demands a reliable preoperative exclusion of malignancy
Information foraging with an oracle
No full textDuring ecological decisions, such as when foraging for food or selecting a weekend activity, we often have to balance the costs and benefits of exploiting known options versus exploring novel ones. Here, we ask how individuals address such cost-benefit tradeoffs during tasks in which we can either explore by ourselves or seek external advice from an oracle (e.g., a domain expert or recommendation system). To answer this question, we designed two studies in which participants chose between inquiring (at a cost) for expert advice from an oracle, or to search for options without guidance, under manipulations affecting the optimal choice. We found that participants showed a greater propensity to seek expert advice when it was instrumental to increase payoff (study A), and when it reduced choice uncertainty, above and beyond payoff maximization (study B). This latter result was especially apparent in participants with greater trait-level intolerance of uncertainty. Taken together, these results suggest that we seek expert advice for both economic goals (i.e., payoff maximization) and epistemic goals (i.e., uncertainty minimization) and that our decisions to ask or not ask for advice are sensitive to cost-benefit tradeoffs
18-Fluorodeoxiglucose positron emission tomography and pancreatic cancer.
No full textLetter to the Edito
TREATMENT OF BUERGER'S DISEASE WITH ELECTRICAL SPINAL CORD STIMULATION. REVIEW OF THREE CASES.
No full text- …
