IBM

About: IBM is a company organization based out in Armonk, New York, United States. It is known for research contribution in the topics: Layer (electronics) & Signal. The organization has 134567 authors who have published 253905 publications receiving 7458795 citations. The organization is also known as: International Business Machines Corporation & Big Blue.

Deep convolutional neural networks for LVCSR

Abstract: Convolutional Neural Networks (CNNs) are an alternative type of neural network that can be used to reduce spectral variations and model spectral correlations which exist in signals. Since speech signals exhibit both of these properties, CNNs are a more effective model for speech compared to Deep Neural Networks (DNNs). In this paper, we explore applying CNNs to large vocabulary speech tasks. First, we determine the appropriate architecture to make CNNs effective compared to DNNs for LVCSR tasks. Specifically, we focus on how many convolutional layers are needed, what is the optimal number of hidden units, what is the best pooling strategy, and the best input feature type for CNNs. We then explore the behavior of neural network features extracted from CNNs on a variety of LVCSR tasks, comparing CNNs to DNNs and GMMs. We find that CNNs offer between a 13-30% relative improvement over GMMs, and a 4-12% relative improvement over DNNs, on a 400-hr Broadcast News and 300-hr Switchboard task.

Universal one-way hash functions and their cryptographic applications

Abstract: We define a Universal One-Way Hash Function family, a new primitive which enables the compression of elements in the function domain. The main property of this primitive is that given an element x. We prove constructively that universal one-way hash functions exist if any 1-1 one-way functions exist.Among the various applications of the primitive is a One-Way based Secure Digital Signature Scheme, a system which is based on the existence of any 1-1 One-Way Functions and is secure against the most general attack known. Previously, all provably secure signature schemes were based on the stronger mathematical assumption that trapdoor one-way functions exist.

Tunable infrared plasmonic devices using graphene/insulator stacks

Abstract: The collective oscillation of carriers--the plasmon--in graphene has many desirable properties, including tunability and low loss. However, in single-layer graphene, the dependence on carrier concentration of both the plasmonic resonance frequency and magnitude is relatively weak, limiting its applications in photonics. Here, we demonstrate transparent photonic devices based on graphene/insulator stacks, which are formed by depositing alternating wafer-scale graphene sheets and thin insulating layers, then patterning them together into photonic-crystal-like structures. We show experimentally that the plasmon in such stacks is unambiguously non-classical. Compared with doping in single-layer graphene, distributing carriers into multiple graphene layers effectively enhances the plasmonic resonance frequency and magnitude, which is different from the effect in a conventional semiconductor superlattice and is a direct consequence of the unique carrier density scaling law of the plasmonic resonance of Dirac fermions. Using patterned graphene/insulator stacks, we demonstrate widely tunable far-infrared notch filters with 8.2 dB rejection ratios and terahertz linear polarizers with 9.5 dB extinction ratios. An unpatterned stack consisting of five graphene layers shields 97.5% of electromagnetic radiation at frequencies below 1.2 THz. This work could lead to the development of transparent mid- and far-infrared photonic devices such as detectors, modulators and three-dimensional metamaterial systems.

On Communicating Finite-State Machines

Abstract: A model of commumcations protocols based on finite-state machines is investigated. The problem addressed is how to ensure certain generally desirable properties, which make protocols \"wellformed,\" that is, specify a response to those and only those events that can actually occur. It is determined to what extent the problem is solvable, and one approach to solving it ts described. Categories and SubJect Descriptors' C 2 2 [Computer-Conununication Networks]: Network Protocols-protocol verification; F 1 1 [Computation by Abstract Devices] Models of Computation--automata; G.2.2 [Discrete Mathematics] Graph Theory--graph algoruhms; trees General Terms: Reliability, Verification Additional

The new dynamics of strategy: Sense-making in a complex and complicated world

Abstract: In this paper, we challenge the universality of three basic assumptions prevalent in organizational decision support and strategy: assumptions of order, of rational choice, and of intent. We describe the Cynefin framework, a sense-making device we have developed to help people make sense of the complexities made visible by the relaxation of these assumptions. The Cynefin framework is derived from several years of action research into the use of narrative and complexity theory in organizational knowledge exchange, decision-making, strategy, and policy-making. The framework is explained, its conceptual underpinnings are outlined, and its use in group sense-making and discourse is described. Finally, the consequences of relaxing the three basic assumptions, using the Cynefin framework as a mechanism, are considered.