Qualcomm

About: Qualcomm is a company organization based out in Farnborough, United Kingdom. It is known for research contribution in the topics: Wireless & Signal. The organization has 19408 authors who have published 38405 publications receiving 804693 citations. The organization is also known as: Qualcomm Incorporated & Qualcomm, Inc..

Learning to cluster in order to transfer across domains and tasks

Abstract: This paper introduces a novel method to perform transfer learning across domains and tasks, formulating it as a problem of learning to cluster. The key insight is that, in addition to features, we can transfer similarity information and this is sufficient to learn a similarity function and clustering network to perform both domain adaptation and cross-task transfer learning. We begin by reducing categorical information to pairwise constraints, which only considers whether two instances belong to the same class or not. This similarity is category-agnostic and can be learned from data in the source domain using a similarity network. We then present two novel approaches for performing transfer learning using this similarity function. First, for unsupervised domain adaptation, we design a new loss function to regularize classification with a constrained clustering loss, hence learning a clustering network with the transferred similarity metric generating the training inputs. Second, for cross-task learning (i.e., unsupervised clustering with unseen categories), we propose a framework to reconstruct and estimate the number of semantic clusters, again using the clustering network. Since the similarity network is noisy, the key is to use a robust clustering algorithm, and we show that our formulation is more robust than the alternative constrained and unconstrained clustering approaches. Using this method, we first show state of the art results for the challenging cross-task problem, applied on Omniglot and ImageNet. Our results show that we can reconstruct semantic clusters with high accuracy. We then evaluate the performance of cross-domain transfer using images from the Office-31 and SVHN-MNIST tasks and present top accuracy on both datasets. Our approach doesn't explicitly deal with domain discrepancy. If we combine with a domain adaptation loss, it shows further improvement.

Complexity-adaptive 2d-to-3d video sequence conversion

Abstract: Techniques for complexity-adaptive and automatic two-dimensional (2D) to three-dimensional (3D) image and video conversion which classifies a frame of a 2D input into one of a flat image class and a non-flat image class are described. The flat image class frame is directly converted into 3D stereo for display. The frame that is classified as a non-flat image class is further processed automatically and adaptively, based on complexity, to create a depth map estimate. Thereafter, the non-flat image class frame is converted into a 3D stereo image using the depth map estimate or an adjusted depth map. The adjusted depth map is processed based on the complexity.

System for delivering relevant user information based on proximity and privacy controls

Abstract: Methods, systems and devices for providing relevant user information to devices within proximity of a user. A wireless identity transmitter associated with the user periodically broadcasts messages that include obscured identifiers. A proximate proximity broadcast receiver may receive and relay the broadcast messages to a central server which may process the included information. Based on categories related to the proximity broadcast receiver and the user, the central server may identify subsets of stored profile information about the user that are relevant to the proximity broadcast receiver. The central server may transmit relevant profile information to devices to assist in activities associated with the proximity broadcast receiver. In an embodiment, the central server may only transit relevant profile information that is authorized by the user via permissions associated with the profile. Further, the central server may transmit payment authentication profile information for use by point-of-sale devices within proximity of the user.

On the design of device-to-device autonomous discovery

Abstract: This paper proposes a synchronous device discovery solution for ad-hoc networks based on the observations that time synchronization, along with an FDM based channel resource allocation, can lead to gains in terms of energy consumption, discovery range, and the number of devices discovered. These attributes are important for the success of proximity-aware networking, where devices autonomously find peer-groups over human mobility scales. In this paper, we develop the PHY and MAC protocols to enable autonomous device discovery. Using both simulations and stochastic-geometry based analysis, we validate our design, and argue that there can be significant gains over a conventional Wi-Fi based solution.

Ldpc encoding methods and apparatus

Abstract: A flexible and relatively hardware efficient LDPC encoder is described. The encoder can be implemented with a level of parallelism which is less than the full parallelism of the code structure used to control the encoding process. Each command of a relatively simple microcode used to describe the code structure can be stored and executed multiple times to complete the encoding of a codeword. Different codeword lengths can be supported using the same set of microcode instructions but with the code being implemented a different number of times depending on the lifting factor selected to be used. The LDPC encoder can switch between encoding codewords of different lengths, without the need to change the stored code description information, by simply changing a code lifting factor used to control the encoding processes. When coding codewords shorter than the maximum supported codeword length some block storage locations and/or registers may go unused.