Showing papers by "Toyota published in 2013"

Stochastic gradient descent with differentially private updates

Abstract: Differential privacy is a recent framework for computation on sensitive data, which has shown considerable promise in the regime of large datasets. Stochastic gradient methods are a popular approach for learning in the data-rich regime because they are computationally tractable and scalable. In this paper, we derive differentially private versions of stochastic gradient descent, and test them empirically. Our results show that standard SGD experiences high variability due to differential privacy, but a moderate increase in the batch size can improve performance significantly.

Systematic exploration of the mechanism of chemical reactions: the global reaction route mapping (GRRM) strategy using the ADDF and AFIR methods

Abstract: Global reaction route mapping (GRRM), a fully-automated search for all important reaction pathways relevant to a given purpose, on the basis of quantum chemical calculations enables systematic elucidation of complex chemical reaction mechanisms However, GRRM had previously been limited to very simple systems This is mainly because such calculations are highly demanding even in small systems when a brute-force sampling is considered Hence, we have developed two independent but complementary methods: anharmonic downward distortion following (ADDF) and artificial force induced reaction (AFIR) methods ADDF can follow reaction pathways starting from local minima on the potential energy surface (PES) toward transition structures (TSs) and dissociation channels AFIR can find pathways starting from two or more reactants toward TSs for their associative reactions In other words, ADDF searches for A → X type isomerization and A → X + Y type dissociation pathways, whereas AFIR finds A + B → X (+ Y) type associative pathways Both follow special paths called the ADDF path and the AFIR path, and these tend to pass through near TSs of corresponding reaction pathways, giving approximate TSs Such approximate TSs can easily be re-optimized to corresponding true TSs by standard geometry optimizations On the basis of these two methods, we have proposed practical strategies of GRRM The GRRM strategies have been applied to a variety of chemical systems ranging from thermal- and photochemical-reactions in small systems to organometallic- and enzyme-catalysis, on the basis of quantum chemical calculations In this perspective, we present an overview of the GRRM strategies and some results of applications Their practical usage for systematic prediction is also discussed

Directive Contrast Based Multimodal Medical Image Fusion in NSCT Domain

Abstract: Multimodal medical image fusion, as a powerful tool for the clinical applications, has developed with the advent of various imaging modalities in medical imaging. The main motivation is to capture most relevant information from sources into a single output, which plays an important role in medical diagnosis. In this paper, a novel fusion framework is proposed for multimodal medical images based on non-subsampled contourlet transform (NSCT). The source medical images are first transformed by NSCT followed by combining low- and high-frequency components. Two different fusion rules based on phase congruency and directive contrast are proposed and used to fuse low- and high-frequency coefficients. Finally, the fused image is constructed by the inverse NSCT with all composite coefficients. Experimental results and comparative study show that the proposed fusion framework provides an effective way to enable more accurate analysis of multimodality images. Further, the applicability of the proposed framework is carried out by the three clinical examples of persons affected with Alzheimer, subacute stroke and recurrent tumor.

Reversible Oxygen Participation to the Redox Processes Revealed for Li1.20Mn0.54Co0.13Ni0.13O2

Abstract: Materials prepared by chemical Li deintercalation with NO2BF4 from Li1.20Mn0.54Co0.13Ni0.13O2 and chemical Li reinsertion with LiI show very similar chemical composition, oxidation state of each transition metal ion, structural properties and electrochemical performance to those of the material recovered after the 1st electrochemical cycle. Investigations combining redox titration, magnetic measurement, neutron diffraction and chemical analyzes reveal that uncommon redox processes are involved during the first charge at high voltage and explain the charge overcapacity and large reversible discharge capacity obtained for this material. This further assesses our proposal that oxygen, in addition to nickel and cobalt, participates to the redox processes in charge: within the bulk oxygen is oxidized without oxygen loss, whereas at the surface oxygen is oxidized to O2 and irreversibly lost from the structure. During the subsequent discharge, in addition to nickel, cobalt and oxygen, manganese is also slightly involved in the redox processes (reduction) to compensate for the initial surface oxygen loss.

Recent Progress in Advanced Materials for Lithium Ion Batteries

Abstract: The development and commercialization of lithium ion batteries is rooted in material discovery. Promising new materials with high energy density are required for achieving the goal toward alternative forms of transportation. Over the past decade, significant progress and effort has been made in developing the new generation of Li-ion battery materials. In the review, I will focus on the recent advance of tin- and silicon-based anode materials. Additionally, new polyoxyanion cathodes, such as phosphates and silicates as cathode materials, will also be discussed.

Cooperative Collision Avoidance at Intersections: Algorithms and Experiments

Abstract: In this paper, we leverage vehicle-to-vehicle (V2V) communication technology to implement computationally efficient decentralized algorithms for two-vehicle cooperative collision avoidance at intersections. Our algorithms employ formal control theoretic methods to guarantee a collision-free (safe) system, whereas overrides are only applied when necessary to prevent a crash. Model uncertainty and communication delays are explicitly accounted for by the model and by the state estimation algorithm. The main contribution of this work is to provide an experimental validation of our method on two instrumented vehicles engaged in an intersection collision avoidance scenario in a test track.

Memory effect in a lithium-ion battery

Abstract: Memory effects are well known to users of nickel-cadmium and nickel-metal-hydride batteries. If these batteries are recharged repeatedly after being only partially discharged, they gradually lose usable capacity owing to a reduced working voltage. Lithium-ion batteries, in contrast, are considered to have no memory effect. Here we report a memory effect in LiFePO4-one of the materials used for the positive electrode in Li-ion batteries-that appears already after only one cycle of partial charge and discharge. We characterize this memory effect of LiFePO4 and explain its connection to the particle-by-particle charge/discharge model. This effect is important for most battery uses, as the slight voltage change it causes can lead to substantial miscalculations in estimating the state of charge of batteries.

A Highly Efficient Mononuclear Iridium Complex Photocatalyst for CO2 Reduction under Visible Light

Abstract: Development of photocatalysts for the reduction of CO2 by sunlight is increasingly becoming an important research area owing to fossil-fuel shortage and global warming. Developing a photosynthetic system that generates solar fuel from CO2, H2O, and sunlight is a promising approach. Photocatalytic systems, including transition-metal complexes such as ruthenium(II) polypyridine carbonyl complexes, cobalt(II) trisbipyridine, and cobalt(II) macrocycles combined with a photosensitizer, can reduce CO2 with a relatively high quantum yield and high product selectivity. Among them, the rhenium(I) bipyridine (bpy) complex systems are the only mononuclear systems that exhibit definite photocatalytic activity for CO2 reduction. A typical example is fac-[Re(bpy)(CO)3Cl], developed by Lehn, which reduces CO2 to CO under UV irradiation without any additional photosensitizers. Cobalt porphyrins can also act as a CO2 reduction catalyst without a photosensitizer. [3c] A fac-[Re(bpy)(CO)3{P(OEt)3}] + complex is an efficient photocatalyst for CO2 reduction in a homogeneous system that selectively produces CO with a quantum yield of 0.38 at the ultraviolet light irradiation of 365 nm. However, the compound must be modified to allow effective use of solar energy because its absorption in the visible region is limited to wavelengths less than 440 nm. Thus, activation of highly active Re complex photocatalysts toward the visible region is necessary. Furthermore, the photocatalytic activity for CO2 reduction is very low in the presence of H2O, even at a concentration of 10 %. Therefore, for CO2 reduction, the development of metal complex photocatalysts that operate under visible light irradiation, even in the presence of H2O, is desirable. In Ir complexes, the stronger spin–orbit coupling coordinates with singlet and triplet excited states, leading to efficient luminescence and visible-light absorption from the singlet–triplet transition. Therefore, Ir complexes have been used as an emitter for electroluminescence devices, a photosensitizer for photocatalytic reactions, and a light-absorber for Gr tzel solar cells. Recently, it was reported that an Ir complex acted as a water oxidation catalyst with a sacrificial electron accepter and a CO2 reduction catalyst with an electronical bias or in the presence of hydrogen. Although Ir complexes are considered suitable for photocatalysis owing to visible-light absorption from S–T transitions and a longer lifetime of the excited state, no studies on Ir complex photocatalysts for CO2 reduction have been reported. This report describes the successful development of a novel photocatalyst, mononuclear iridium(III) terpyridine (tpy) 2-phenylpyridine (ppy) complex [Ir(tpy)(ppy)Cl] ([Irppy]), which selectively reduced CO2 to CO under visible light at 480 nm without additional photosensitizers such as in the case for Re complexes. Furthermore, advantages of the Ir complexes over Re complexes include: 1) greater photocatalytic activity for CO2 reduction; 2) CO2 reduction under visible light, such as at a wavelength of 480 nm; and 3) the photocatalytic activity is maintained (including selectivity) even in a solution containing H2O. [Ir-ppy] catalyzed the reduction of CO2 molecules to CO under visible-light irradiation. Figure 1 shows the photocatalytic formation of CO over [Ir-ppy] compared with conventional [Re(bpy)(CO)3Cl] under visible light irradiation

LIMERIC: A Linear Adaptive Message Rate Algorithm for DSRC Congestion Control

Abstract: Wireless vehicle-to-vehicle (V2V) and vehicle-toinfrastructure (V2I) communication holds great promise for significantly reducing the human and financial costs of vehicle collisions A common characteristic of this communication is the broadcast of a device's core state information at regular intervals (eg, vehicle speed and location or traffic signal state and timing) Unless controlled, the aggregate of these broadcasts will congest the channel under dense traffic scenarios, reducing the effectiveness of collision avoidance applications that use transmitted information Active congestion control using distributed techniques is a topic of great interest for establishing the scalability of this technology This paper defines a new adaptive congestion control algorithm that can be applied to the message rate of devices in this vehicular environment While other published approaches rely on binary control, the LInear MEssage Rate Integrated Control (LIMERIC) algorithm takes advantage of full-precision control inputs that are available on the wireless channel The result is provable convergence to fair and efficient channel utilization in the deterministic environment, under simple criteria for setting adaptive parameters This “perfect” convergence avoids the limit cycle behavior that is inherent to binary control We also discuss several practical aspects associated with implementing LIMERIC, including guidelines for the choice of system parameters to obtain desired utilization outcomes, a gain saturation technique that maintains robust convergence under all conditions, convergence with asynchronous updates, and using channel load to determine the aggregate message rate that is observable at a receiver This paper also extends the convergence analysis for two important cases, ie, measurement noise in the input signal and delay in the update process This paper illustrates key analytical results using MATLAB numerical results and employs standard NS-2 simulations to demonstrate the performance of LIMERIC in several high-density scenarios

Solar CO2 reduction using H2O by a semiconductor/metal-complex hybrid photocatalyst: enhanced efficiency and demonstration of a wireless system using SrTiO3 photoanodes

Abstract: Solar formate production from CO2 and H2O was achieved with no external electrical bias by combining an InP/[RuCP] semiconductor/metal-complex hybrid photocathode with a reduced SrTiO3 (r-STO) photoanode. The conversion efficiency from solar to chemical energy was improved from 0.03 to 0.14% compared to a previous system utilizing a TiO2 photoanode. Stimulated electron transfer from the photoanode to the photocathode is the main cause for the observed improvement, due to an enlarged difference in the band-energy position between r-STO and InP. Since r-STO showed high H2O oxidation selectivity even in the presence of formate, a r-STO/InP/[RuCP] wireless device successfully performed solar CO2 reduction in a one-compartment reactor with no proton exchange membrane, yielding a solar conversion efficiency of 0.08%.

Vehicle control device

Abstract: An ECU (800) includes an abnormality determination unit (810), a condition determination unit (820) and a running control unit (830). The abnormality determination unit (810) determines whether or not there is an abnormality (shift pattern abnormality) in a combination of shift signals provided from a shift position sensor. After occurrence of the shift pattern abnormality is determined, the condition determination unit (820) determines whether or not a first condition that a shift pattern changes into a normal pattern of a drive range is satisfied, and whether or not a second condition that a vehicle is running (a driver has an intention to cause the vehicle to run) is satisfied. When occurrence of the shift pattern abnormality is determined, the running control unit (830) stops generation of driving force of the vehicle and prohibits switching to the drive range. When the condition determination unit (820) determines that the above-mentioned first condition and second condition are satisfied after occurrence of the shift pattern abnormality is determined, the running control unit (830) switches the shift range to the drive range indicated by the shift pattern and recovers the driving force.

Efficient Classification for Additive Kernel SVMs

Abstract: We show that a class of nonlinear kernel SVMs admits approximate classifiers with runtime and memory complexity that is independent of the number of support vectors. This class of kernels, which we refer to as additive kernels, includes widely used kernels for histogram-based image comparison like intersection and chi-squared kernels. Additive kernel SVMs can offer significant improvements in accuracy over linear SVMs on a wide variety of tasks while having the same runtime, making them practical for large-scale recognition or real-time detection tasks. We present experiments on a variety of datasets, including the INRIA person, Daimler-Chrysler pedestrians, UIUC Cars, Caltech-101, MNIST, and USPS digits, to demonstrate the effectiveness of our method for efficient evaluation of SVMs with additive kernels. Since its introduction, our method has become integral to various state-of-the-art systems for PASCAL VOC object detection/image classification, ImageNet Challenge, TRECVID, etc. The techniques we propose can also be applied to settings where evaluation of weighted additive kernels is required, which include kernelized versions of PCA, LDA, regression, k-means, as well as speeding up the inner loop of SVM classifier training algorithms.

LED and CMOS Image Sensor Based Optical Wireless Communication System for Automotive Applications

Abstract: An optical wireless communication (OWC) system based on a light-emitting-diode (LED) transmitter and a camera receiver has been developed for use in the automotive area. The automotive OWC system will require Mb/s-class data rates and the ability to quickly detect LEDs from an image. The key to achieving this is improvements to the capabilities of the image sensor mounted on the camera receiver. In this paper, we report on a novel OWC system equipped with an optical communication image sensor (OCI), which is newly developed using CMOS technology. To obtain higher transmission rates, the OCI employs a specialized “communication pixel (CPx)” capable of responding promptly to optical intensity variations. Furthermore, a new quick LED detection technique, based on a 1-bit flag image which only reacts to high-intensity objects, is formulated. The communication pixels, ordinary image pixels, and associated circuits (including 1-bit flag image output circuits) are then integrated into the OCI. This paper describes the design, fabrication, and capabilities of the OCI, as well as the development of the LED and image sensor based OWC system, which boasts a 20-Mb/s/pixel data rate without LED detection and a 15-Mb/s/pixel data rate with a 16.6-ms real-time LED detection.

Signal Processing and Machine Learning with Differential Privacy: Algorithms and Challenges for Continuous Data

Abstract: Private companies, government entities, and institutions such as hospitals routinely gather vast amounts of digitized personal information about the individuals who are their customers, clients, or patients. Much of this information is private or sensitive, and a key technological challenge for the future is how to design systems and processing techniques for drawing inferences from this large-scale data while maintaining the privacy and security of the data and individual identities. Individuals are often willing to share data, especially for purposes such as public health, but they expect that their identity or the fact of their participation will not be disclosed. In recent years, there have been a number of privacy models and privacy-preserving data analysis algorithms to answer these challenges. In this article, we will describe the progress made on differentially private machine learning and signal processing.

Wireless Charging System Using Secure Wireless Charging Protocols

Abstract: The disclosure includes a system and method for charging a target object wirelessly. The system includes a processor and a memory storing instructions that when executed cause the system to: receive data describing a charging request from a target object; generate a challenge responsive to the charging request; send the challenge to the target object; receive a response from the target object; verify the response to determine that the response matches the challenge and a first set of secret data shared with a tagging device; determine that a location associated with the target object satisfies a safe charging range responsive to the verification of the response; and instruct a power transmitter associated with the target object to transmit power wirelessly to a power receiver associated with the target object responsive to the verification of the response and the determination that the location satisfies the safe charging range.

A 100-m Range 10-Frame/s 340 $\,\times\, $ 96-Pixel Time-of-Flight Depth Sensor in 0.18- $\mu\hbox{m}$ CMOS

Abstract: This paper introduces a single-photon detection technique for time-of-flight distance ranging based on the temporal and spatial correlation of photons. A proof-of-concept prototype achieving depth imaging up to 100 meters with a resolution of 340 × 96 pixels at 10 frames/s was implemented. At the core of the system, a sensor chip comprising 32 macro-pixels based on an array of single-photon avalanche diodes featuring an optical fill factor of 70% was fabricated in a 0.18-μm CMOS. The chip also comprises an array of 32 circuits capable of generating precise triggers upon correlation events as well as of sampling the number of photons involved in each correlation event, and an array of 32 12-b time-to-digital converters. Characterization of the TDC array led to -0.52 LSB and 0.73 LSB of differential and integral nonlinearities, respectively. Quantitative evaluation of the TOF sensor under strong solar background light, i.e., 80 klux, revealed a repeatability error better than 10 cm throughout the distance range of 100 m, thus leading to a relative precision of 0.1%. In the same condition, the relative nonlinearity error was 0.37%. In order to show the suitability of our approach in a real-world situation, experimental results in which the depth sensor was operated in a typical traffic scenario are also reported.

Layered perovskite oxide: a reversible air electrode for oxygen evolution/reduction in rechargeable metal-air batteries.

Abstract: For the development of a rechargeable metal-air battery, which is expected to become one of the most widely used batteries in the future, slow kinetics of discharging and charging reactions at the air electrode, i.e., oxygen reduction reaction (ORR) and oxygen evolution reaction (OER), respectively, are the most critical problems. Here we report that Ruddlesden–Popper-type layered perovskite, RP-LaSr3Fe3O10 (n = 3), functions as a reversible air electrode catalyst for both ORR and OER at an equilibrium potential of 1.23 V with almost no overpotentials. The function of RP-LaSr3Fe3O10 as an ORR catalyst was confirmed by using an alkaline fuel cell composed of Pd/LaSr3Fe3O10–2x(OH)2x·H2O/RP-LaSr3Fe3O10 as an open circuit voltage (OCV) of 1.23 V was obtained. RP-LaSr3Fe3O10 also catalyzed OER at an equilibrium potential of 1.23 V with almost no overpotentials. Reversible ORR and OER are achieved because of the easily removable oxygen present in RP-LaSr3Fe3O10. Thus, RP-LaSr3Fe3O10 minimizes efficiency losses ...

Inter-brain synchronization during coordination of speech rhythm in human-to-human social interaction

Abstract: Behavioral rhythms synchronize between humans for communication; however, the relationship of brain rhythm synchronization during speech rhythm synchronization between individuals remains unclear. Here, we conducted alternating speech tasks in which two subjects alternately pronounced letters of the alphabet during hyperscanning electroencephalography. Twenty pairs of subjects performed the task before and after each subject individually performed the task with a machine that pronounced letters at almost constant intervals. Speech rhythms were more likely to become synchronized in human–human tasks than human–machine tasks. Moreover, theta/alpha (6–12 Hz) amplitudes synchronized in the same temporal and lateral-parietal regions in each pair. Behavioral and inter-brain synchronizations were enhanced after human–machine tasks. These results indicate that inter-brain synchronizations are tightly linked to speech synchronizations between subjects. Furthermore, theta/alpha inter-brain synchronizations were also found in subjects while they observed human–machine tasks, which suggests that the inter-brain synchronization might reflect empathy for others' speech rhythms.

Tracking via Robust Multi-task Multi-view Joint Sparse Representation

Abstract: Combining multiple observation views has proven beneficial for tracking. In this paper, we cast tracking as a novel multi-task multi-view sparse learning problem and exploit the cues from multiple views including various types of visual features, such as intensity, color, and edge, where each feature observation can be sparsely represented by a linear combination of atoms from an adaptive feature dictionary. The proposed method is integrated in a particle filter framework where every view in each particle is regarded as an individual task. We jointly consider the underlying relationship between tasks across different views and different particles, and tackle it in a unified robust multi-task formulation. In addition, to capture the frequently emerging outlier tasks, we decompose the representation matrix to two collaborative components which enable a more robust and accurate approximation. We show that the proposed formulation can be efficiently solved using the Accelerated Proximal Gradient method with a small number of closed-form updates. The presented tracker is implemented using four types of features and is tested on numerous benchmark video sequences. Both the qualitative and quantitative results demonstrate the superior performance of the proposed approach compared to several state-of-the-art trackers.

Hidden Factor Analysis for Age Invariant Face Recognition

Abstract: Age invariant face recognition has received increasing attention due to its great potential in real world applications. In spite of the great progress in face recognition techniques, reliably recognizing faces across ages remains a difficult task. The facial appearance of a person changes substantially over time, resulting in significant intra-class variations. Hence, the key to tackle this problem is to separate the variation caused by aging from the person-specific features that are stable. Specifically, we propose a new method, called Hidden Factor Analysis (HFA). This method captures the intuition above through a probabilistic model with two latent factors: an identity factor that is age-invariant and an age factor affected by the aging process. Then, the observed appearance can be modeled as a combination of the components generated based on these factors. We also develop a learning algorithm that jointly estimates the latent factors and the model parameters using an EM procedure. Extensive experiments on two well-known public domain face aging datasets: MORPH (the largest public face aging database) and FGNET, clearly show that the proposed method achieves notable improvement over state-of-the-art algorithms.

First-Principles Study of Alkali and Alkaline Earth Ion Intercalation in Iron Hexacyanoferrate: The Important Role of Ionic Radius

Abstract: Hexacyanoferrate compounds recently have attracted much interest because of their potential as aqueous and nonaqueous battery cathodes in the application of large-scale energy storage. Here, the feasibility to use iron hexacyanoferrate, Fe[Fe(CN)6], as a cathode candidate for nonaqueous rechargeable batteries is assessed with first-principle calculations. The structural deformation induced by the intercalation of alkali (Li+, Na+, K+, Rb+, Cs+) and alkaline earth (Mg2+, Ca2+, Sr2+, Ba2+) ions into Fe[Fe(CN)6] induces is negligible as compared to other Li-ion battery cathodes. The intercalation is strongly affected by the ionic radius of inserted species. With increasing ionic size, the most stable interstitial site changes from face-centered site to body-centered site. More importantly, the voltages for the intercalation are highly correlated to the ionic radius of the inserted species. The insertion of cations with larger ionic radius happens at higher voltages, and thus provides higher energy density. H...

From N to N+1: Multiclass Transfer Incremental Learning

Abstract: Since the seminal work of Thrun [16], the learning to learn paradigm has been defined as the ability of an agent to improve its performance at each task with experience, with the number of tasks. Within the object categorization domain, the visual learning community has actively declined this paradigm in the transfer learning setting. Almost all proposed methods focus on category detection problems, addressing how to learn a new target class from few samples by leveraging over the known source. But if one thinks of learning over multiple tasks, there is a need for multiclass transfer learning algorithms able to exploit previous source knowledge when learning a new class, while at the same time optimizing their overall performance. This is an open challenge for existing transfer learning algorithms. The contribution of this paper is a discriminative method that addresses this issue, based on a Least-Squares Support Vector Machine formulation. Our approach is designed to balance between transferring to the new class and preserving what has already been learned on the source models. Extensive experiments on subsets of publicly available datasets prove the effectiveness of our approach.

Corrosion of magnesium electrolytes: chlorides – the culprit

Abstract: Chloride containing magnesium electrolytes are corrosive towards non noble metals. Currently the development of non-corrosive magnesium electrolytes is a key challenge on the road to a rechargeable magnesium battery. The component responsible for corrosion of magnesium electrolytes has not been previously elucidated. Here we clarify that chlorides in the cation (Mg2(μ-Cl)3·6THF)+ are a major culprit for corrosion. We also corroborate the feasibility of ion exchange reactions as a suitable synthetic approach towards magnesium electrolytes which do not contain the cation (Mg2(μ-Cl)3·6THF)+. Our results indicate that magnesium organoborates are an interesting class of magnesium electrolytes which undergo magnesium deposition and dissolution and are non-corrosive in nature at high voltages.