We challenge the longstanding assumption that the mean-field approximation for variational inference in Bayesian neural networks is severely restrictive. We argue mathematically that full-covariance approximations only improve the ELBO if they improve the expected log-likelihood. We further show that deeper mean-field networks are able to express predictive distributions approximately equivalent to shallower full-covariance networks. We validate these observations empirically, demonstrating that deeper models decrease the divergence between diagonal- and full-covariance Gaussian fits to the true posterior.

The prevalence of wearable sensors (e.g., smart wristband) is enabling an unprecedented opportunity to not only inform health and wellness states of individuals, but also assess and infer demographic information and personality. This can allow us a deeper personalized insight beyond how many steps we took or what is our heart rate. However, before we can achieve this goal of personalized insight about an individual, we have to resolve a number of shortcomings: 1) wearable-sensory time series is often of variable-length and incomplete due to different data collection periods (e.g., wearing behavior varies by person); 2) inter-individual variability to external factors like stress and environment. This paper addresses these challenges and brings us closer to the potential of personalized insights whether about health or personality or job performance about an individual by developing a novel representation learning algorithm, HeartSpace. Specifically, HeartSpace is capable of encoding time series data with variable-length and missing values via the integration of a time series encoding module and a pattern aggregation network. Additionally, HeartSpace implements a Siamese-triplet network to optimize representations by jointly capturing intra- and inter-series correlations during the embedding learning process. Our empirical evaluation over two different data presents significant performance gains over state-of-the-art baselines in a variety of applications, including personality prediction, demographics inference, user identification.

The machine learning community has become increasingly interested in the energy efficiency of neural networks. The Spiking Neural Network (SNN) is a promising approach to energy-efficient computing, since its activation levels are quantized into temporally sparse, one-bit values (i.e., "spike" events), which additionally converts the sum over weight-activity products into a simple addition of weights (one weight for each spike). However, the goal of maintaining state-of-the-art (SotA) accuracy when converting a non-spiking network into an SNN has remained an elusive challenge, primarily due to spikes having only a single bit of precision. Adopting tools from signal processing, we cast neural activation functions as quantizers with temporally-diffused error, and then train networks while smoothly interpolating between the non-spiking and spiking regimes. We apply this technique to the Legendre Memory Unit (LMU) to obtain the first known example of a hybrid SNN outperforming SotA recurrent architectures---including the LSTM, GRU, and NRU---in accuracy, while reducing activities to at most 3.74 bits on average with 1.26 significant bits multiplying each weight. We discuss how these methods can significantly improve the energy efficiency of neural networks.

Heretofore, neural networks with external memory are restricted to single memory with lossy representations of memory interactions. A rich representation of relationships between memory pieces urges a high-order and segregated relational memory. In this paper, we propose to separate the storage of individual experiences (item memory) and their occurring relationships (relational memory). The idea is implemented through a novel Self-attentive Associative Memory (SAM) operator. Found upon outer product, SAM forms a set of associative memories that represent the hypothetical high-order relationships between arbitrary pairs of memory elements, through which a relational memory is constructed from an item memory. The two memories are wired into a single sequential model capable of both memorization and relational reasoning. We achieve competitive results with our proposed two-memory model in a diversity of machine learning tasks, from challenging synthetic problems to practical testbeds such as geometry, graph, reinforcement learning, and question answering.

This paper presents a generative approach to speech enhancement based on a recurrent variational autoencoder (RVAE). The deep generative speech model is trained using clean speech signals only, and it is combined with a nonnegative matrix factorization noise model for speech enhancement. We propose a variational expectation-maximization algorithm where the encoder of the RVAE is fine-tuned at test time, to approximate the distribution of the latent variables given the noisy speech observations. Compared with previous approaches based on feed-forward fully-connected architectures, the proposed recurrent deep generative speech model induces a posterior temporal dynamic over the latent variables, which is shown to improve the speech enhancement results.

Try it out: Christopher Hesse provided the live TensorFlow demonstration for pix2pix. You can "translate" your doodles of cats, facades, and other things into "photo-realistic" images. It was surely way more than just funny sketch translation: with pre-defined settings, you can transform an aerial photo to a map, a daylight photo to a night view etc. Conditional Adversarial Networks detect the patterns and translate them to the demanded topics (you have to define your target image task). Networks are trained on specific labeled image datasets. NVidia brought this method to another level with GauGAN -- one of the experiments in their AI Playground. You can drive sketches using segmentation: every color is applied to a specific object or material.

He was a co-recipient of the 2018 ACM A.M. Turing Award for his work in deep learning.

Yoshua Bengio: Yoshua Bengio OCFRSC (born 1964 in Paris, France) is a Canadian computer scientist, most noted for his work on artificial neural networks and deep learning.[1][2][3] He was a co-recipient of the 2018 ACM A.M. Turing Award for his work in deep learning.[4] He is a professor at the Department of Computer Science and Operations Research at the Université de Montréal and scientific director of the Montreal Institute for Learning Algorithms (MILA). Geoffrey Hinton: Geoffrey Everest HintonCCFRSFRSC[11] (born 6 December 1947) is an English Canadian cognitive psychologist and computer scientist, most noted for his work on artificial neural networks. Since 2013 he divides his time working for Google (Google Brain) and the University of Toronto.

Quality control can decrease variations in the performance of colonoscopists and improve the effectiveness of colonoscopy to prevent colorectal cancers. Unfortunately, routine quality control is difficult to carry out because a practical method is lacking. The aim of this study was to develop an automatic quality control system (AQCS) and assess whether it could improve polyp and adenoma detection in clinical practice. First, we developed AQCS based on deep convolutional neural network models for timing of the withdrawal phase, supervising withdrawal stability, evaluating bowel preparation, and detecting colorectal polyps. Next, consecutive patients were prospectively randomized to undergo routine colonoscopies with or without the assistance of AQCS.

There are numerous applications that combine data collected from sensors with machine-learning based classification models to predict the type of event or objects observed. Both the collection of the data itself and the classification models can be tuned for optimal performance, but we hypothesize that additional gains can be realized by jointly assessing both factors together. Through this research, we used a seismic event dataset and two neural network classification models that issued probabilistic predictions on each event to determine whether it was an earthquake or a quarry blast. Real world applications will have constraints on data collection, perhaps inmore » terms of a budget for the number of sensors or on where, when, or how data can be collected. We compare different methods of determining the set of sensors in each subnetwork in terms of their predictive accuracy and the number of events that they observe overall.