Voice conversion has gained increasing popularity in many applications of speech synthesis. The idea is to change the voice identity from one speaker into another while keeping the linguistic content unchanged. Many voice conversion approaches rely on the use of a vocoder to reconstruct the speech from acoustic features, and as a consequence, the speech quality heavily depends on such a vocoder. In this paper, we propose NVC-Net, an end-to-end adversarial network, which performs voice conversion directly on the raw audio waveform of arbitrary length. By disentangling the speaker identity from the speech content, NVC-Net is able to perform non-parallel traditional many-to-many voice conversion as well as zero-shot voice conversion from a short utterance of an unseen target speaker. Importantly, NVC-Net is non-autoregressive and fully convolutional, achieving fast inference. Our model is capable of producing samples at a rate of more than 3600 kHz on an NVIDIA V100 GPU, being orders of magnitude faster than state-of-the-art methods under the same hardware configurations. Objective and subjective evaluations on non-parallel many-to-many voice conversion tasks show that NVC-Net obtains competitive results with significantly fewer parameters.

Recent advances in large-scale pre-training such as GPT-3 allow seemingly high quality text to be generated from a given prompt. However, such generation systems often suffer from problems of hallucinated facts, and are not inherently designed to incorporate useful external information. Grounded generation models appear to offer remedies, but their training typically relies on rarely-available parallel data where corresponding information-relevant documents are provided for context. We propose a framework that alleviates this data constraint by jointly training a grounded generator and document retriever on the language model signal. The model learns to reward retrieval of the documents with the highest utility in generation, and attentively combines them using a Mixture-of-Experts (MoE) ensemble to generate follow-on text. We demonstrate that both generator and retriever can take advantage of this joint training and work synergistically to produce more informative and relevant text in both prose and dialogue generation.

The use of deep features coming from pre-trained neural networks for unsupervised anomaly detection purposes has recently gathered momentum in the computer vision field. In particular, industrial inspection applications can take advantage of such features, as demonstrated by the multiple successes of related methods on the MVTec Anomaly Detection (MVTec AD) dataset. These methods make use of neural networks pre-trained on auxiliary classification tasks such as ImageNet. However, to our knowledge, no comparative study of robustness to the low data regimes between these approaches has been conducted yet. For quality inspection applications, the handling of limited sample sizes may be crucial as large quantities of images are not available for small series. In this work, we aim to compare three approaches based on deep pre-trained features when varying the quantity of available data in MVTec AD: KNN, Mahalanobis, and PaDiM. We show that although these methods are mostly robust to small sample sizes, they still can benefit greatly from using data augmentation in the original image space, which allows to deal with very small production runs.

We study calibration in question answering, estimating whether model correctly predicts answer for each question. Unlike prior work which mainly rely on the model's confidence score, our calibrator incorporates information about the input example (e.g., question and the evidence context). Together with data augmentation via back translation, our simple approach achieves 5-10% gains in calibration accuracy on reading comprehension benchmarks. Furthermore, we present the first calibration study in the open retrieval setting, comparing the calibration accuracy of retrieval-based span prediction models and answer generation models. Here again, our approach shows consistent gains over calibrators relying on the model confidence. Our simple and efficient calibrator can be easily adapted to many tasks and model architectures, showing robust gains in all settings.

While Moore's law has driven exponential computing power expectations, its nearing end calls for new avenues for improving the overall system performance. One of these avenues is the exploration of new alternative brain-inspired computing architectures that promise to achieve the flexibility and computational efficiency of biological neural processing systems. Within this context, neuromorphic intelligence represents a paradigm shift in computing based on the implementation of spiking neural network architectures tightly co-locating processing and memory. In this paper, we provide a comprehensive overview of the field, highlighting the different levels of granularity present in existing silicon implementations, comparing approaches that aim at replicating natural intelligence (bottom-up) versus those that aim at solving practical artificial intelligence applications (top-down), and assessing the benefits of the different circuit design styles used to achieve these goals. First, we present the analog, mixed-signal and digital circuit design styles, identifying the boundary between processing and memory through time multiplexing, in-memory computation and novel devices. Next, we highlight the key tradeoffs for each of the bottom-up and top-down approaches, survey their silicon implementations, and carry out detailed comparative analyses to extract design guidelines. Finally, we identify both necessary synergies and missing elements required to achieve a competitive advantage for neuromorphic edge computing over conventional machine-learning accelerators, and outline the key elements for a framework toward neuromorphic intelligence.

Neural models have shown impressive performance gains in answering queries from natural language text. However, existing works are unable to support database queries, such as "List/Count all female athletes who were born in 20th century", which require reasoning over sets of relevant facts with operations such as join, filtering and aggregation. We show that while state-of-the-art transformer models perform very well for small databases, they exhibit limitations in processing noisy data, numerical operations, and queries that aggregate facts. We propose a modular architecture to answer these database-style queries over multiple spans from text and aggregating these at scale. We evaluate the architecture using WikiNLDB, a novel dataset for exploring such queries. Our architecture scales to databases containing thousands of facts whereas contemporary models are limited by how many facts can be encoded. In direct comparison on small databases, our approach increases overall answer accuracy from 85% to 90%. On larger databases, our approach retains its accuracy whereas transformer baselines could not encode the context.

The field of natural language understanding has experienced exponential progress in the last few years, with impressive results in several tasks. This success has motivated researchers to study the underlying knowledge encoded by these models. Despite this, attempts to understand their semantic capabilities have not been successful, often leading to non-conclusive, or contradictory conclusions among different works. Via a probing classifier, we extract the underlying knowledge graph of nine of the most influential language models of the last years, including word embeddings, text generators, and context encoders. This probe is based on concept relatedness, grounded on WordNet. Our results reveal that all the models encode this knowledge, but suffer from several inaccuracies. Furthermore, we show that the different architectures and training strategies lead to different model biases. We conduct a systematic evaluation to discover specific factors that explain why some concepts are challenging. We hope our insights will motivate the development of models that capture concepts more precisely.

Given a graph or similarity matrix, we consider the problem of recovering a notion of true distance between the nodes, and so their true positions. Through new insights into the manifold geometry underlying a generic latent position model, we show that this can be accomplished in two steps: matrix factorisation, followed by nonlinear dimension reduction. This combination is effective because the point cloud obtained in the first step lives close to a manifold in which latent distance is encoded as geodesic distance. Hence, a nonlinear dimension reduction tool, approximating geodesic distance, can recover the latent positions, up to a simple transformation. We give a detailed account of the case where spectral embedding is used, followed by Isomap, and provide encouraging experimental evidence for other combinations of techniques.

Online fraud and identity theft is a global problem that has only been exacerbated with increased online transactions amid the COVID-19 pandemic. In particular, it is estimated that Brazilian companies lose over $41 billion due to fraud every year. In an attempt to tackle this problem head on, Lincoln Ando and Raphael Melo started idwall in mid-2016. São Paulo-based idwall started as an automated background check solution and has since grown into a suite of data and identity validation and risk analysis products. For the consumer market, its "MeuID" app is aimed at users who want to change the way they identify themselves and share their data.

Transformer model architectures have become an indispensable staple in deep learning lately for their effectiveness across a range of tasks. Recently, a surge of "X-former" models have been proposed which improve upon the original Transformer architecture. However, most of these variants make changes only around the quadratic time and memory complexity of self-attention, i.e. the dot product between the query and the key. What's more, they are calculate solely in Euclidean space. In this work, we propose a novel Transformer with Hyperbolic Geometry (THG) model, which take the advantage of both Euclidean space and Hyperbolic space. THG makes improvements in linear transformations of self-attention, which are applied on the input sequence to get the query and the key, with the proposed hyperbolic linear. Extensive experiments on sequence labeling task, machine reading comprehension task and classification task demonstrate the effectiveness and generalizability of our model. It also demonstrates THG could alleviate overfitting.