Searcharxiv⌕ Search

arXiv subjects

Leonardo Neves

Publications and source records attributed to Leonardo Neves.

45 records · Page 3Linked to original sources

Ptychography of pure quantum states

Ptychography is an imaging technique in which a localized illumination scans overlapping regions of an object and generates a set of diffraction intensities used to computationally reconstruct its complex-valued transmission function. We propose a quantum analogue of this technique designed to reconstruct $d$-dimensional pure states. A set of $n$ rank-$r$ projectors "scans" overlapping parts of an input state and the moduli of the $d$ Fourier amplitudes of each part are measured. These $nd$ outcomes are fed into an iterative phase retrieval algorithm that estimates the state. Using $d$ up to 100 and $r$ around $d/2$, we performed numerical simulations for single systems in an economic ($n=4$) and a costly ($n=d$) scenario, as well as for multiqubit systems ($n=6\log d$). This numeric study included realistic amounts of depolarization and poissonian noise, and all scenarios yielded, in general, reconstructions with infidelities below $10^{-2}$. The method is shown, therefore, to be resilient to noise and, for any $d$, requires a simple and fast postprocessing algorithm. We show that the algorithm is equivalent to an alternating gradient search, which ensures that it does not suffer from local-minima stagnation. Unlike traditional approaches to state reconstruction, the ptychographic scheme uses a single measurement basis; the diversity and redundancy in the measured data---key for its success---are provided by the overlapping projections. We illustrate the simplicity of this scheme with the paradigmatic multiport interferometer.

quant-ph↗

Train One Get One Free: Partially Supervised Neural Network for Bug Report Duplicate Detection and Clustering

Tracking user reported bugs requires considerable engineering effort in going through many repetitive reports and assigning them to the correct teams. This paper proposes a neural architecture that can jointly (1) detect if two bug reports are duplicates, and (2) aggregate them into latent topics. Leveraging the assumption that learning the topic of a bug is a sub-task for detecting duplicates, we design a loss function that can jointly perform both tasks but needs supervision for only duplicate classification, achieving topic clustering in an unsupervised fashion. We use a two-step attention module that uses self-attention for topic clustering and conditional attention for duplicate detection. We study the characteristics of two types of real world datasets that have been marked for duplicate bugs by engineers and by non-technical annotators. The results demonstrate that our model not only can outperform state-of-the-art methods for duplicate classification on both cases, but can also learn meaningful latent clusters without additional supervision.

cs.CL↗

Multimodal Named Entity Recognition for Short Social Media Posts

We introduce a new task called Multimodal Named Entity Recognition (MNER) for noisy user-generated data such as tweets or Snapchat captions, which comprise short text with accompanying images. These social media posts often come in inconsistent or incomplete syntax and lexical notations with very limited surrounding textual contexts, bringing significant challenges for NER. To this end, we create a new dataset for MNER called SnapCaptions (Snapchat image-caption pairs submitted to public and crowd-sourced stories with fully annotated named entities). We then build upon the state-of-the-art Bi-LSTM word/character based NER models with 1) a deep image network which incorporates relevant visual context to augment textual information, and 2) a generic modality-attention module which learns to attenuate irrelevant modalities while amplifying the most informative ones to extract contexts from, adaptive to each sample and token. The proposed MNER model with modality attention significantly outperforms the state-of-the-art text-only NER models by successfully leveraging provided visual contexts, opening up potential applications of MNER on myriads of social media platforms.

cs.CL↗

Visual Features for Context-Aware Speech Recognition

Automatic transcriptions of consumer-generated multi-media content such as "Youtube" videos still exhibit high word error rates. Such data typically occupies a very broad domain, has been recorded in challenging conditions, with cheap hardware and a focus on the visual modality, and may have been post-processed or edited. In this paper, we extend our earlier work on adapting the acoustic model of a DNN-based speech recognition system to an RNN language model and show how both can be adapted to the objects and scenes that can be automatically detected in the video. We are working on a corpus of "how-to" videos from the web, and the idea is that an object that can be seen ("car"), or a scene that is being detected ("kitchen") can be used to condition both models on the "context" of the recording, thereby reducing perplexity and improving transcription. We achieve good improvements in both cases and compare and analyze the respective reductions in word error rate. We expect that our results can be used for any type of speech processing in which "context" information is available, for example in robotics, man-machine interaction, or when indexing large audio-visual archives, and should ultimately help to bring together the "video-to-text" and "speech-to-text" communities.

cs.CL↗

Proposal for Automated Operations for Single-Photon Multipath Qudits

We propose a method for implementing automated state transformations on single-photon multipath qudits encoded in a one-dimensional transverse spatial domain. It relies on transferring the encoding from this domain to the orthogonal one by applying a spatial phase modulation with diffraction gratings, merging all the initial propagation paths with a stable interferometric network, and filtering out the unwanted diffraction orders. The automated feature is attained by utilizing a programmable phase-only spatial light modulator (SLM) where properly designed diffraction gratings displayed on its screen will implement the desired transformations, including, among others, projections, permutations and random operations. We discuss the losses in the process which is, in general, inherently nonunitary. Some examples of transformations are presented and, considering a realistic scenario, we analyse how they will be affected by the pixelated structure of the SLM screen. The method proposed here enables one to implement much more general transformations on multipath qudits than it is possible with an SLM alone operating in the diagonal basis of which-path states. Therefore, it will extend the range of applicability for this encoding in high-dimensional quantum information and computing protocols as well as fundamental studies in quantum theory.

quant-ph↗

Optimal probabilistic dense coding schemes

Dense coding with non-maximally entangled states has been investigated in many different scenarios. We revisit this problem for protocols adopting the standard encoding scheme. In this case, the set of possible classical messages cannot be perfectly distinguished due to the non-orthogonality of the quantum states carrying them. So far, the decoding process has been approached in two ways: (i) The message is always inferred, but with an associated (minimum) error; (ii) the message is inferred without error, but only sometimes; in case of failure, nothing else is done. Here, we generalize on these approaches and propose novel optimal probabilistic decoding schemes. The first uses quantum-state separation to increase the distinguishability of the messages with an optimal success probability. This scheme is shown to include (i) and (ii) as special cases and continuously interpolate between them, which enables the decoder to trade-off between the level of confidence desired to identify the received messages and the success probability for doing so. The second scheme, called multistage decoding, applies only for qudits ($d$-level quantum systems with $d>2$) and consists of further attempts in the state identification process in case of failure in the first one. We show that this scheme is advantageous over (ii) as it increases the mutual information between the sender and receiver.

quant-ph↗

Hybrid photonic entanglement: Realization, characterization and applications

We show that the quantum disentanglement eraser implemented on a two-photon system from parametric down-conversion is a general method to create hybrid photonic entanglement, namely the entanglement between different degrees of freedom of the photon pair. To demonstrate this, we generate and characterize a source with tunable degree of hybrid entanglement between two qubits, one encoded in the transverse momentum and position of a photon, and the other in the polarization of its partner. In addition, we show that a simple extension of our setup enables the generation of two-photon qubit-qudit hybrid entangled states. Finally, we discuss the advantages that this type of entanglement can bring for an optical quantum network.

quant-ph↗

Propagation of spatially entangled qudits through free space

We show the propagation of entangled states of high-dimensional quantum systems. The qudits states were generated using the transverse correlation of the twin photons produced by spontaneous parametric down-conversion. Their free-space distribution was performed at the laboratory scale and the propagated states maintained a high-fidelity with their original form. The use of entangled qudits allow an increase in the quantity of information that can be transmitted and may also guarantee more privacy for communicating parties. Therefore, studies about propagating entangled states of qudits are important for the effort of building quantum communication networks.

quant-ph↗

Generation of maximally entangled states of qudits using twin photons

We report an experiment to generate maximally entangled states of D-dimensional quantum systems, qudits, by using transverse spatial correlations of two parametric down-converted photons. Apertures with D-slits in the arms of the twin fotons define the qudit space. By manipulating the pump beam correctly the twin photons will pass only by symmetrically opposite slits, generating entangled states between these differents paths. Experimental results for qudits with D=4 and D=8 are shown. We demonstrate that the generated states are entangled states.

quant-ph↗