In this paper we show how to train statistical machine translation systems on reallife tasks using only non-parallel monolingual data from two languages. We present a modification of the method shown in (Ravi and Knight, 2011) that is scalable to vocabulary sizes of several thousand words. On the task shown in (Ravi and Knight, 2011) we obtain better results with only 5 % of the computational effort when running our method with an n-gram language model. The efficiency improvement of our method allows us to run experiments with vocabulary sizes of around 5,000 words, such as a non-parallel version of the VERBMOBIL corpus. We also report results using data from the monolingual French and English GIGAWORD corpora. 1

We present a simple objective function that when optimized yields accurate solutions to both decipherment and cognate pair identification problems. The objective simultaneously scores a matching between two alphabets and a matching between two lexicons, each in a different language. We introduce a simple coordinate descent procedure that efficiently finds effective solutions to the resulting combinatorial optimization problem. Our system requires only a list of words in both languages as input, yet it competes with and surpasses several state-of-the-art systems that are both substantially more complex and make use of more information. 1

Smart water meters will soon provide real-time access to instantaneous water usage in many homes, and disaggregation is the problem of deciding how much of that usage is due to individual fixtures in the home. Existing disaggregation techniques require additional water sensing infrastructure and/or a manual characterization of each water fixture, which can be expensive and time consuming. In this paper, we describe a novel technique called WaterSense to perform fixture-level disaggregation using only a handful of inexpensive motion sensors. WaterSense automatically infers how many fixtures are in each room, and how much water each fixture uses. We evaluate the system using a 7-day in-situ evaluation in 2 diverse multi-resident homes with a total of 10 different water fixtures and 467 fixture events and observe that our approach achieves 86 % classification accuracy in identifying individual water fixture events and 80-90 % accuracy in determining the water consumption of individual water fixtures.

ABSTRACT When open addressing IS used to resolve collisions in a hash table, a given set of keys may be arranged in many ways, typically this depends on the order in which the keys are inserted It is shown that arrangements minimizing either the average or worst-case number of probes required to retrieve any key in the table can be found using an algorithm for the assignment problem. The worst-case retrieval time can be reduced to O(log2(M)) with probablhty 1- e(M) when storing Mkeys In a table of size M, where ~(M) ~ 0 as M ~ ~ We also examine insertion algorithms to see how to apply these ideas for a dynamically changing set of keys

Abstract—In this paper we show how the capacity of the uplink of a multiuser system can be increased by a scheduling strategy, which pairs the transmission of users in different time/frequency/code slots according to the channel quality. The optimal scheduling strategy is equivalent to a combinatorial optimization problem. We show how this problem can be solved efficiently by using the Hungarian method. We then show that, by using the proposed scheduling scheme, the performance of Minimum Mean Square Error detection approaches the one of Maximum Likelihood detection, as the number of users increases. Index terms — Capacity, Scheduling, multiuser uplink, MIMO, assignment problem, Hungarian method.

We introduce a method to learn a mixture of submodular “shells ” in a large-margin setting. A submodular shell is an abstract submodular function that can be instantiated with a ground set and a set of parameters to produce a submodular function. A mixture of such shells can then also be so instantiated to produce a more complex submodular function. What our algorithm learns are the mixture weights over such shells. We provide a risk bound guarantee when learning in a large-margin structured-prediction setting using a projected subgradient method when only approximate submodular optimization is possible (such as with submodular function maximization). We apply this method to the problem of multi-document summarization and produce the best results reported so far on the widely used NIST DUC-05 through DUC-07 document summarization corpora. 1

. Content-based image retrieval systems allow the user to interactively search image databases looking for those images which are similar to a specified query image. To thisend, region-based systems decompose each database image into a set of "homogeneous" regions. Similarity between images is then assessed by computing similarity between regions and combining the results at image level. In this paper we propose a correct approach to region matching maximizing the overall similarity score between images. The presented approach only relies on the assumption that regions' scores are to be combined using a monotonic function. Experimental results obtained using an existing system show the e#ectiveness of our approach with respect to existing region matching heuristics.Then, to reduce query costs, we present an index-based algorithm that can make use of any distance-based access structure, and demonstrate its e#ciency on a medium-size image data-set. 1

We study the weighted circuit constraint in the context of constraint programming. It appears as a substructure in many practical applications, particularly routing problems. We propose a domain filtering algorithm for the weighted circuit constraint that is based on the 1-tree relaxation of Held and Karp. In addition, we study domain filtering based on an additive bounding procedure that combines the 1-tree relaxation with the assignment problem relaxation. Experimental results on Traveling Salesman Problem instances demonstrate that our filtering algorithms can dramatically reduce the problem size. In particular, the search tree size and solving time can be reduced by several orders of magnitude, compared to existing constraint programming approaches. Moreover, for medium-size problem instances, our method is competitive with the state-of-the-art special-purpose TSP solver Concorde.

Abstract—Control systems for Atomic Force Microscopes (AFMs) tend to be specific to a particular model of device, and further have a tendency to require that they be written to target an inconvenient execution environment. This paper addresses these problems by describing a high-level programming system for an AFM in which the device-specific low level code has been separated into a different process accessible across the network. This frees the bulk of the code from the assorted constraints imposed by the specific device, and also allows for the insertion of an abstraction layer between the high level control code and the device itself, making it possible to write device independent control code.

Abstract — We consider the problem of multi-robot taskallocation when robots have to deal with uncertain utility estimates. Typically an allocation is performed to maximize expected utility; we consider a means for measuring the robustness of a given optimal allocation when robots have some measure of the uncertainty (e.g., a probability distribution, or moments of such distributions). We introduce a new O(n 4) algorithm, the Interval Hungarian algorithm, that extends the classic Kuhn-Munkres Hungarian algorithm to compute the maximum interval of deviation (for each entry in the assignment matrix) which will retain the same optimal assignment. This provides an efficient measurement of the tolerance of the allocation to the uncertainties, for both a specific interval and a set of interrelated intervals. We conduct experiments both in simulation and with physical robots to validate the approach and to gain insight into the effect of location uncertainty on allocations for multi-robot multi-target navigation tasks. I.