We have demonstrated nonlinear cross-phase modulation in electro-optic crystals using intense, single-cycle terahertz (THz) radiation. Individual THz pulses, generated by coherent transition radiation emitted by subpicosecond electron bunches, have peak energies of up to 100 microJ per pulse. The time-dependent electric field of the intense THz pulses induces cross-phase modulation in electro-optic crystals through the Pockels effect, leading to spectral shifting, broadening, and modulation of copropagating laser pulses. The observed THz-induced cross-phase modulation agrees well with a time-dependent phase-shift model.

The purpose of this paper is to detect pedestrians from images. This paper proposes a method for extracting feature descriptors consisting of co-occurrence histograms of oriented gradients (CoHOG). Including co-occurrence with various positional offsets, the feature descriptors can express complex shapes of objects with local and global distributions of gradient orientations. Our method is evaluated with a simple linear classifier on two famous pedestrian detection benchmark datasets: "DaimlerChrysler pedestrian classification benchmark dataset" and "INRIA person data set". The results show that proposed method reduces miss rate by half compared with HOG, and outperforms the state-of-the-art methods on both datasets.

Constituent parsing is typically modeled by a chart-based algorithm under probabilistic context-free grammars or by a transition-based algorithm with rich features. Previous models rely heavily on richer syntactic information through lexicalizing rules, splitting categories, or memorizing long histories. However enriched models incur numerous parameters and sparsity issues, and are insufficient for capturing various syntactic phenomena. We propose a neural network structure that explicitly models the unbounded history of actions performed on the stack and queue employed in transition-based parsing, in addition to the representations of partially parsed tree structure. Our transition-based neural constituent parsing achieves performance comparable to the state-of-the-art parsers, demonstrating F1 score of 90.68% for English and 84.33% for Chinese, without reranking, feature templates or additional data to train model parameters.

The 13th harmonic of a Ti:sapphire (Ti:S) laser in the plateau region was injected as a seeding source to a 250-MeV free-electron-laser (FEL) amplifier. When the amplification conditions were fulfilled, strong enhancement of the radiation intensity by a factor of 650 was observed. The random and uncontrollable spikes, which appeared in the spectra of the Self-Amplified Spontaneous Emission (SASE) based FEL radiation without the seeding source, were found to be suppressed drastically to form to a narrow-band, single peak profile at 61.2 nm. The properties of the seeded FEL radiation were well reproduced by numerical simulations. We discuss the future precept of the seeded FEL scheme to the shorter wavelength region.

Watanabe et al. Reply: After carefully evaluating the preceding Comment by Bonafacio et al. [1] on our Letter [2], we concluded that those comments do not affect our experimental results, nor alter our analysis and the conclusions presented in our Letter. The following is the detailed response to the comments. In the following the detailed response to the comments are given in the order of remark (2), (3), and (1), followed by our conclusion.In our single-pass high-gain free-electron laser (FEL) amplifier experiments, the seed pulse is first amplified in the exponential gain regime where the pulse lengthens; after saturation of the exponential gain regime the FEL transitions into the superradiant regime. Superradiance in a FEL amplifier manifests itself in several ways in addition to the radiation power scaling as I 2 . In the superradiant regime the FEL amplifier pulse length shrinks, the spectral bandwidth increases, and the FEL pulse energy continues to grow (without the use of any undulator tapering) (Refs. [8, 11, 13, 18] of [2]). In our Letter the continuous growth of the FEL energy in the nonlinear regime (Fig. 2), the FEL pulse shortening by roughly 50% (Fig. 3), and the resulting spectral broadening (Fig. 3) were all observed and the results were consistent with the numerical simulations (Fig. 4). These simultaneous observations provide compelling proof of superradiance in our FEL amplifier experiments.The measurement of the power scaling as I 2 as suggested in remark (2) is very challenging for our experimental setup, as a change in the beam charge would affect the electron beam emittance, energy spread, bunch length, and the bunch current distribution, especially in the photoinjector and in the compressor. This means that the FEL saturated power and the distance where the FEL radiation evolves as a superradiant pulse are changed accordingly. For the above reasons, in our experimental conditions a simple scaling as P / I 2 is not expected. Note that measuring the FEL energy as a function of the undulator length has been widely used to characterize single-pass high-gain FELs [2 -5, 23].In [18] it is shown that after saturation the pulse propagates at a constant velocity approximately equal to c and scaling laws are derived in the variable z, defined as the bunch coordinate along the undulator. In response to remark (3) we show that it is straightforward to demonstrate under this assumption the equivalence with the scaling laws derived in Refs. [8, 12]. Here we indicate with (z l ; t l ) the time and position coordinates in the laboratory. In [18] the equations are expressed in a coordinate system (; z) moving at the constant velocity z c in the longitudinal direction.

Measuring domain relevance of data and identifying or selecting well-fit domain data for machine translation (MT) is a well-studied topic, but denoising is not yet. Denoising is concerned with a different type of data quality and tries to reduce the negative impact of data noise on MT training, in particular, neural MT (NMT) training. This paper generalizes methods for measuring and selecting data for domain MT and applies them to denoising NMT training. The proposed approach uses trusted data and a denoising curriculum realized by online data selection. Intrinsic and extrinsic evaluations of the approach show its significant effectiveness for NMT to train on data with severe noise.

Spatial and energy distributions of energetic electrons produced by an ultrashort, intense laser pulse with a short focal length optical system (Ti:sapphire, 12 TW, 50 fs, lambda=790 nm, f/3.5) in a He gas jet are measured. They are shown to depend strongly on the contrast ratio and shape of the laser prepulse. The wave breaking of the plasma waves at the front of the shock wave formed by a proper laser prepulse is found to make a narrow-cone (0.1pi mm mrad) electron injection. These electrons are further accelerated by the plasma wake field generated by the laser pulse up to tens of MeV forming a Maxwell-like energy distribution. In the case of nonmonotonic prepulse, hydrodynamic instability at the shock front leads to a broader, spotted spatial distribution. The numerical analysis based on a two-dimensional (2D) hydrodynamic (for the laser prepulse) and 2D particle-in-cell (PIC) simulation justifies the mechanism of electron acceleration. The PIC calculation predicts that electrons with energy from 10 to 40 MeV form a bunch with a pulse duration of about 40 fs.

In statistical machine translation, the generation of a translation hypothesis is computationally expensive. If arbitrary reorderings are permitted, the search problem is NP-hard. On the other hand, if we restrict the possible reorderings in an appropriate way, we obtain a polynomial-time search algorithm. We investigate different reordering constraints for phrase-based statistical machine translation, namely the IBM constraints and the ITG constraints. We present efficient dynamic programming algorithms for both constraints. We evaluate the constraints with respect to translation quality on two Japanese-English tasks. We show that the reordering constraints improve translation quality compared to an unconstrained search that permits arbitrary phrase reorderings. The ITG constraints preform best on both tasks and yield statistically significant improvements compared to the unconstrained search.