--In this paper, we propose a model for the Environment Sound Classification T ask (ESC) that consists of multiple feature channels given as input to a Deep Convolutional Neural Network (CNN). The novelty of the paper lies in using multiple feature channels consisting of Mel-Frequency Cepstral Coefficients (MFCC), Gammatone Frequency Cepstral Coefficients (GFCC), the Constant Q-transform (CQT) and Chromagram. Such multiple features have never been used before for signal or audio processing. Also, we employ a deeper CNN (DCNN) compared to previous models, consisting of 2D separable convolutions working on time and feature domain separately. The model also consists of max pooling layers that downsample time and feature domain separately. We use some data augmentation techniques to further boost performance. Our model is able to achieve state-of- the-art performance on all three benchmark environment sound classification datasets, i.e. the UrbanSound8K (97.35%), T o the best of our knowledge, this is the first time that a single environment sound classification model is able to achieve state-of-the-art results on all three datasets. For ESC-10 and ESC-50 datasets, the accuracy achieved by the proposed model is beyond human accuracy of 95.7% and 81.3% respectively. I NTRODUCTION T HERE are many important applications related to speech and audio processing. One of the most important application is the Environment Sound Classification (ESC) that deals with distinguishing between sounds from the real environment. It is a complex task that involves classifying a sound event into an appropriate class such as siren, dog barking, airplane, people talking etc. This task is quite different compared to Automatic Speech Recognition (ASR) [1], since environment sound features differ drastically from speech sounds. In ASR, speech is converted to text. However, in ESC, there is no such thing as speech, just sounds. So, ESC models are quite different compared to ASR models.

Deep neural networks have been increasingly used in software engineering and program analysis tasks. They usually take a program and make some predictions about it, e.g., bug prediction. We call these models neural program analyzers. The reliability of neural programs can impact the reliability of the encompassing analyses. In this paper, we describe our ongoing efforts to develop effective techniques for testing neural programs. We discuss the challenges involved in developing such tools and our future plans. In our preliminary experiment on a neural model recently proposed in the literature, we found that the model is very brittle, and simple perturbations in the input can cause the model to make mistakes in its prediction.

Deep Neural Networks (DNNs) have achieved state-of-the-art accuracy performance in many tasks. However, recent works have pointed out that the outputs provided by these models are not well-calibrated, seriously limiting their use in critical decision scenarios. In this work, we propose to use a decoupled Bayesian stage, implemented with a Bayesian Neural Network (BNN), to map the uncalibrated probabilities provided by a DNN to calibrated ones, consistently improving calibration. Our results evidence that incorporating uncertainty provides more reliable probabilistic models, a critical condition for achieving good calibration. We report a generous collection of experimental results using high-accuracy DNNs in standardized image classification benchmarks, showing the good performance, flexibility and robust behavior of our approach with respect to several state-of-the-art calibration methods. Code for reproducibility is provided.

We investigate the combination of actor-critic reinforcement learning algorithms with uniform large-scale experience replay and propose solutions for two challenges: (a) efficient actor-critic learning with experience replay (b) stability of very off-policy learning. We employ those insights to accelerate hyper-parameter sweeps in which all participating agents run concurrently and share their experience via a common replay module. To this end we analyze the bias-variance tradeoffs in V-trace, a form of importance sampling for actor-critic methods. Based on our analysis, we then argue for mixing experience sampled from replay with on-policy experience, and propose a new trust region scheme that scales effectively to data distributions where V-trace becomes unstable. We provide extensive empirical validation of the proposed solution. We further show the benefits of this setup by demonstrating state-of-the-art data efficiency on Atari among agents trained up until 200M environment frames.

We propose Dynamically Pruned Message Passing Networks (DPMPN) for large-scale knowledge graph reasoning. In contrast to existing models, embedding-based or path-based, we learn an input-dependent subgraph to explicitly model a sequential reasoning process. Each subgraph is dynamically constructed, expanding itself selectively under a flow-style attention mechanism. In this way, we can not only construct graphical explanations to interpret prediction, but also prune message passing in Graph Neural Networks (GNNs) to scale with the size of graphs. We take the inspiration from the consciousness prior proposed by Bengio to design a two-GNN framework to encode global input-invariant graph-structured representation and learn local input-dependent one coordinated by an attention module. Experiments show the reasoning capability in our model that is providing a clear graphical explanation as well as predicting results accurately, outperforming most state-of-the-art methods in knowledge base completion tasks.

An increasingly popular approach to alleviate this issue is to first learn general language representations on unlabeled data, which are then integrated in task-specific downstream systems. This approach was first popularized by word embeddings (Mikolov et al., 2013b; This work was performed during an internship at Facebook AI Research. Pennington et al., 2014), but has recently been superseded by sentence-level representations (Peters et al., 2018; Devlin et al., 2019). Nevertheless, all these works learn a separate model for each language and are thus unable to leverage information across different languages, greatly limiting their potential performance for low-resource languages. In this work, we are interested in universal language agnostic sentence embeddings, that is, vector representations of sentences that are general with respect to two dimensions: the input language and the NLP task.

TEACH a man how to fish, and you can feed him for a lifetime. Homegrown artificial intelligence (AI) company Xjera Labs aims to do just that, with the aid of neural networks and deep learning algorithms. Its AI engine Athena can train software to perform specific tasks, whether these pertain to recognising sushi, teaching drones to fly or spotting intruders. The company's co-founder and chief executive Ethan Chu compares Athena to a school for its "pupils" - in Xjera's case, these are its products named XIntelligence, XTransport and XHound. Taken together, the trio offers a bevy of functions, including facial recognition, emotional analysis and identifying vehicle types.

The authors of this paper are from Google. They present a neural network for generating raw audio waves. Their model is fully probabilistic and autoregressive, and it generates state-of-the-art text-to-speech results for both English and Mandarin. WaveNet is an audio generative model based on the PixelCNN. In this generative model, each audio sample is conditioned on the previous audio sample.

Fujifilm SonoSite Inc. and the Allen Institute of Artificial Intelligence (AI2) Incubator, builder of AI-first startups, announced a collaboration to interpret ultrasound images with AI, enabling new ultrasound applications and enhanced accuracy. Fujifilm SonoSite has enlisted assistance from the AI2 Incubator to deploy deep learning models on portable ultrasound products. Together, the AI2 Incubator and Fujifilm SonoSite will work to improve image analysis, allowing for the interpretation of a much wider range of ultrasound scenarios. Within the field of medical imaging, deep learning-based techniques have brought breakthroughs across a wide range of scenarios, including detecting tuberculosis (TB) in X-ray scans and diagnosing metastatic breast cancer in pathology slides. Compared to other modalities such as X-ray, computed tomography (CT) and positron emission tomography (PET), ultrasound is more affordable, portable and does not expose patients to ionizing radiation.

The presented work demonstrates the training of recurrent neural networks (RNNs) from distributions of atom coordinates in solid state structures that were obtained using ab initio molecular dynamics (AIMD) simulations. AIMD simulations on solid state structures are treated as a multi-variate time-series problem. By referring interactions between atoms over the simulation time to temporary correlations among them, RNNs find patterns in the multi-variate time-dependent data, which enable forecasting trajectory paths and potential energy profiles. Two types of RNNs, namely gated recurrent unit and long short-term memory networks, are considered. The model is described and compared against a baseline AIMD simulation on an iridium oxide slab. Findings demonstrate that both networks can potentially be harnessed for accelerated statistical sampling in computational materials research.