Xerox Corporation

Naver Corporation

Conduent Business Services, LLC

Naver Labs Corporation

Conduent Business Solutions, LLC

Christopher R Dance

Mauritius Seeger

Stuart Taylor

Onno R Zoeter

Marco Bressan

William M Newman

Guillaume M Bouchard

Gabriela Csurka

Julien Perez

Tomi Silander

Ercan E Kuruoglu

David N Snowdon

Alain Franciosa

Jiang Duan

Victor Ciriza

Nicola Cancedda

Jutta Willamowski

Lixin Fan

Eric Gaussier

Michael J Flynn

Venkatesh Guru Rao

Yu-An Sun

Laurent Donini

Alex S Taylor

Shengbo Guo

Damian Arregui

Hervé Dejean

Tamás Gaál

Guy J Ruddock

Roberto Ramon Rodriguez

Michel Aractingi

Theo Cachet

Mihajlo Grbovic

Madalina Barbaiani

Guoping Qiu

Szilárd Zsolt Fazekas

Kimberly Moravec

Guoping Qui

A computer-implemented method for performing few-shot imitation is disclosed. The method comprises obtaining at least one set of training data, wherein each set of training data is associated with a task and comprises (i) one of samples of rewards and a reward function, (ii) one of samples of state transitions and a transition distribution, and (iii) a set of first demonstrations, training a policy network embodied in an agent using reinforcement learning by inputting at least one set of first demonstrations of the at least one set of training data into the policy network, and by maximizing a risk measure or an average return over the at least one set of first demonstrations of the at least one set of training data based on respective one or more reward functions or respective samples of rewards, obtaining a set of second demonstrations associated with a new task, and inputting the set of second demonstrations and an observation of a state into the trained policy network for performing the new task.

Demonstration-conditioned reinforcement learning for few-shot imitation

A training system for training a trained model for use by a navigating robot to perform visual navigation includes memory including N base virtual training environments, each of the N base virtual training environments including a field of view at a location within an indoor space, where N is an integer greater than 1. A randomization module is configured to generate N varied virtual training environments based on the N base virtual training environments, respectively, by varying at least one characteristic of the respective N base virtual training environments. A training module is configured to train the trained model for use by the navigating robot to perform visual navigation based on a training set including: the N base virtual training environments; and the N varied virtual training environments.

Systems and methods for improving generalization in visual navigation

In a method for monitoring a tracked parameter for a set of targets, a first methodology is applied during a current time interval to acquire observations of the tracked parameter for a selected sub-set of the set of targets. A second methodology is applied during the current time interval to acquire observations of the tracked parameter for at least those targets that are not in the selected sub-set of the set of targets. A variance forecast is generated for an upcoming time interval for each target based on a current variance for the target and on either observations acquired using the first methodology or using the second methodology. The selected sub-set of the set of targets is updated for the upcoming time interval to maximize an aggregation of variance forecasts over the updated selected sub-set of the set of targets.

Multi-target tracking for demand management

Data characterizing a system is received at an electronic processor. For example, parking event data from parking sensors of a parking facility is received. The electronic processor constructs a current state for the system (e.g. parking occupancy state of the parking facility) at a current time from the received data. State probabilities at a future time are computed (e.g. occupancy state probabilities are computed for the parking facility) using a continuous-time Markov chain model modified by multiplying the time input to the model by a random variable and scaling the state probabilities by an expectation of the random variable. In parking occupancy forecasting, parking guidance information is generated based at least on the computed occupancy state probabilities, and is transmitted to an electronic device other than the electronic processor (e.g. a parking recommendation transmitted to a vehicle navigation device, or a control signal transmitted to a 'lot full' sign).

Forecasting with matrix powers

A system and method for generating an occupancy model are disclosed. The model is learned using occupancy data for zones, each zone including cells, which are occupied or not at a given time, each with a sensor, which may be reporting or not. The data provides an observed occupancy corresponding to a number of cells in the respective zone which have reporting sensors, and the number of those sensors which are reporting that the respective cell is occupied. The occupancy model is based on a demand model and a sensor noise model which accounts for behavior of the non-reporting sensors. The noise model assumes that the probability of a sensor being in the reporting state is dependent on whether the respective cell is occupied or not. The model can fit the occupancy data better than one which assumes that non-reporting cells are occupied with the same frequency as reporting ones.

System and method for generating an occupancy model

An alignment method includes, for a source sentence in a source language, identifying whether the sentence includes at least one candidate term comprising a contiguous subsequence of words of the source sentence. A target sentence in a target language is aligned with the source sentence. This includes developing a probabilistic model which models conditional probability distributions for alignments between words of the source sentence and words of the target sentence and generating an optimal alignment based on the probabilistic model, including, where the source sentence includes the at least one candidate term, enforcing a contiguity constraint which requires that all the words of the target sentence which are aligned with an identified candidate term form a contiguous subsequence of the target sentence.

Method for aligning sentences at the word level enforcing selective contiguity constraints

A method and apparatus for paying for an existing report are provided. In the method, an existing report for which a first structure of entities is identified is received. Thereafter, a set of other reports for which respective second structures of entities are identified is received. A weighting for other reports in the set of other reports is assigned, based on the respective structures. A payment for the existing report is extracted based on the weighting, a selected scoring rule, and the set of other reports.

Method for estimation of a payment for an existing report based on subsequent reports which provides incentives for reporters to report truthfully

A system and method are provided. The method includes establishing a current belief about the multidimensional distribution of buyers' valuations for at least one item, and, based on the current belief, proposing at least one pricing mechanism, each pricing mechanism establishing a price for the at least one item. Observed buyers' responses to at least one of the set of proposed pricing mechanisms are stored. Region censored updates to the belief about the multidimensional distribution of buyers' valuations are conducted, based on the observed responses, to generate a new belief about the multidimensional distribution of buyers' valuations. Based on the new belief, a pricing mechanism establishing a price for the at least one item is proposed, that is expected to improve a seller's welfare under the new belief, relative to the originally proposed mechanism or mechanisms.

Multi-dimensional price determination

A system and method for conducting a lottery for at least one item are provided. The method includes, for each of a plurality of buyers, receiving a buyer's declared valuation for each of at least one item being offered in a lottery by a seller, the item having an assigned non-deterministic probability of being allocated to the buyer, providing insured prices for outcomes of the lottery which are a function of the buyer's declared valuation of the at least one item, randomly drawing an allocation of each of the at least one item to a respective one of the buyers, based on its assigned non-deterministic probability, and allocating the insured prices to the buyers based on respective outcomes of the random drawing.

Limited lottery insurance

According to various embodiments, the present teachings include inventory control policies that are defined in terms of functions of aggregate cost rates, involving thresholds &#x3a9; and an order-up-to point S. An embodiment of the present teachings includes a method. The method includes tracking an inventory position of each of the plurality of items by a logistics network and determining an item cost rate for each of the plurality of items based on the tracked inventory position. The method also includes determining an aggregate cost rate for the plurality of items based on the determined item cost rates, comparing the aggregate cost rate with a cost rate threshold &#x3a9;, and ordering the plurality of items to an order-up-to point S if the compared aggregate cost rate is greater than or equal to the cost rate threshold &#x3a9;.

Methods for supply chain management

In a scheduling or planning process, a hybrid process is determined representing operation of a system. The hybrid process includes stochastic jumps. An approximate linear program is constructed respective to the hybrid process and a set of control actions for controlling the system. The approximate linear program includes a stochastic jump constraint term representative of stochastic jumps of the hybrid process. The approximate linear program is solved. The system is controlled based on the solved approximate linear program.

Solving continuous stochastic jump control problems with approximate linear programming

A valuation optimization method includes generating offeree decision information (buyer decision information, by way of illustrative example) by presenting a sequence of mechanisms to a sequence of offerees wherein the mechanisms comprise menus of transaction offers (sale offer menus, by way of illustrative example). Actual transactions (sale transactions, by way of illustrative example) are conducted responsive to acceptances of transaction offers by buyers. At a selected time in the generating, an offeree valuation distribution belief and the current mechanism are updated optimize an offeree's utility. The offeree's utility comprises an offeree's utility function constrained by a robust learning constraint computed based on a local differential of an earlier offeree's utility function with respect to the earlier offeree's valuation.

Price optimization with robust learning constraint

A system and a method of providing a time-out for a device, such as a printer are provided. The time-out determines when the device is shifted from a higher energy to a lower energy mode, absent the arrival of another job to be processed by the device. The method includes acquiring data comprising a set of inter-arrival times for at least one device over a period of time, such as a week and, for each of a set of candidate time-outs, deriving a probability from the data that an inter-arrival time from the set of inter-arrival times is greater than the candidate time-out. A cost function is computed, based on the derived probability and a robustness term which allows adversarial action not predicted by the histogram to be taken into account. A time-out for the at least one device can then be identified for which the cost function is a minimum.

Printer time-out

A method for image enhancement includes providing for a semantic class to be assigned to a digital image based on image content, the assigned semantic class being selected from a plurality of semantic classes. The method further includes providing for an aesthetic enhancement to be applied to the image based on image quality of the image and the assigned semantic class, the enhancement including at least one enhancement dimension selected from a plurality of enhancement dimensions.

Class-based image enhancement system

Growing community of inventors

Grenoble, France

Christopher R Dance