So over the past few months, I've been spending a significant amount of time working with TensorFlow and Keras (with TensorFlow under Keras), and I've been running into the question of when I should use which toolset. For those of you who haven't worked with Keras yet, it's a Python library designed for easy neural network assembly that comes with a large number of prepackaged network types, ranging from convolutional networks in two- and three- dimensional flavors, to long- and short-term networks, to more general recurrent networks. Building networks using Keras is straight-forward -- though determining exactly what network to build isn't. The semantics Keras uses in its API design are very layer-oriented, making network assembly relatively intuitive. TensorFlow, on the other hand, is arguably more powerful -- but doesn't have all the prepackaged networks.

Nearly all of the deep learning based image analysis methods work on reconstructed images, which are obtained from original acquisitions via solving inverse problems. The reconstruction algorithms are designed for human observers, but not necessarily optimized for DNNs. It is desirable to train the DNNs directly from the original data which lie in a different domain with the images. In this work, we proposed an end-to-end DNN for abnormality detection in medical imaging. A DNN was built as the unrolled version of iterative reconstruction algorithms to map the acquisitions to images, and followed by a 3D convolutional neural network (CNN) to detect the abnormality in the reconstructed images. The two networks were trained jointly in order to optimize the entire DNN for the detection task from the original acquisitions. The DNN was implemented for lung nodule detection in low-dose chest CT. The proposed end-to-end DNN demonstrated better sensitivity and accuracy for the task compared to a two-step approach, in which the reconstruction and detection DNNs were trained separately. A significant reduction of false positive rate on suspicious lesions were observed, which is crucial for the known over-diagnosis in low-dose lung CT imaging. The images reconstructed by the proposed end-to-end network also presented enhanced details in the region of interest.

Duan 1,2 1 Center for Quantum Information, IIIS, Tsinghua University, Beijing 100084, PR China 2 Department of Physics, University of Michigan, Ann Arbor, Michigan 48109, USA A central task in the field of quantum computing is to find applications where quantum computer could provide exponential speedup over any classical computer [1-3]. Machine learning represents an important field with broad applications where quantum computer may offer significant speedup [4-8]. Several quantum algorithms for discriminative machine learning [9] have been found based on efficient solving of linear algebraic problems [10-15], with potential exponential speedup in runtime under the assumption of effective input from a quantum random access memory [16]. In machine learning, generative models represent another large class [9] which is widely used for both supervised and unsupervised learning [17, 18]. Here, we propose an efficient quantum algorithm for machine learning based on a quantum generative model. We prove that our proposed model is exponentially more powerful to represent probability distributions compared with classical generative models and has exponential speedup in training and inference at least for some instances under a reasonable assumption in computational complexity theory. Our result opens a new direction for quantum machine learning and offers a remarkable example in which a quantum algorithm shows exponential improvement over any classical algorithm in an important application field. Machine learning and artificial intelligence represent a very important application area which could be revolutionized by quantum computers with clever algorithms that offer exponential speedup [4, 5]. The candidate algorithms with potential exponential speedup so far rely on efficient quantum solution of linear system of equations or linear algebraic problems [12-15]. Those algorithms require quantum random access memory (QRAM) as a critical component in addition to a quantum computer. In a QRAM, the number of required quantum routers scales up exponentially with the number of qubits in those algorithms [16, 19]. This exponential overhead in resource requirement poses a significant challenge for its experimental implementation and is a caveat for fair comparison with corresponding classical algorithms [5, 20]. In this paper, we propose a quantum algorithm with potential exponential speedup for machine learning basedFigure 1: Classical and quantum generative models. A factor graph is a bipartite graph where one group of the vertices represent variables (denoted by circles) and the other group of vertices represent positive functions (denoted by squares) acting on connected variables. The corresponding probability distribution is given by the product of all these functions. Each variable connects to at most a constant number of functions which introduce correlations in the probability distribution.b,

A grand challenge of the 21st century cosmology is to accurately estimate the cosmological parameters of our Universe. A major approach to estimating the cosmological parameters is to use the large-scale matter distribution of the Universe. Galaxy surveys provide the means to map out cosmic large-scale structure in three dimensions. Information about galaxy locations is typically summarized in a "single" function of scale, such as the galaxy correlation function or power-spectrum. We show that it is possible to estimate these cosmological parameters directly from the distribution of matter. This paper presents the application of deep 3D convolutional networks to volumetric representation of dark-matter simulations as well as the results obtained using a recently proposed distribution regression framework, showing that machine learning techniques are comparable to, and can sometimes outperform, maximum-likelihood point estimates using "cosmological models". This opens the way to estimating the parameters of our Universe with higher accuracy.

This paper presents an automated approach for interpretable feature recommendation for solving signal data analytics problems. The method has been tested by performing experiments on datasets in the domain of prognostics where interpretation of features is considered very important. The proposed approach is based on Wide Learning architecture and provides means for interpretation of the recommended features. It is to be noted that such an interpretation is not available with feature learning approaches like Deep Learning (such as Convolutional Neural Network) or feature transformation approaches like Principal Component Analysis. Results show that the feature recommendation and interpretation techniques are quite effective for the problems at hand in terms of performance and drastic reduction in time to develop a solution. It is further shown by an example, how this human-in-loop interpretation system can be used as a prescriptive system.

Calcium imaging permits optical measurement of neural activity. Since intracellular calcium concentration is an indirect measurement of neural activity, computational tools are necessary to infer the true underlying spiking activity from fluorescence measurements. Bayesian model inversion can be used to solve this problem, but typically requires either computationally expensive MCMC sampling, or faster but approximate maximum-a-posteriori optimization. Here, we introduce a flexible algorithmic framework for fast, efficient and accurate extraction of neural spikes from imaging data. Using the framework of variational autoencoders, we propose to amortize inference by training a deep neural network to perform model inversion efficiently. The recognition network is trained to produce samples from the posterior distribution over spike trains. Once trained, performing inference amounts to a fast single forward pass through the network, without the need for iterative optimization or sampling. We show that amortization can be applied flexibly to a wide range of nonlinear generative models and significantly improves upon the state of the art in computation time, while achieving competitive accuracy. Our framework is also able to represent posterior distributions over spike-trains. We demonstrate the generality of our method by proposing the first probabilistic approach for separating backpropagating action potentials from putative synaptic inputs in calcium imaging of dendritic spines.

Compression and computational efficiency in deep learning have become a problem of great significance. In this work, we argue that the most principled and effective way to attack this problem is by adopting a Bayesian point of view, where through sparsity inducing priors we prune large parts of the network. We introduce two novelties in this paper: 1) we use hierarchical priors to prune nodes instead of individual weights, and 2) we use the posterior uncertainties to determine the optimal fixed point precision to encode the weights. Both factors significantly contribute to achieving the state of the art in terms of compression rates, while still staying competitive with methods designed to optimize for speed or energy efficiency.

We present a new autoencoder-type architecture that is trainable in an unsupervised mode, sustains both generation and inference, and has the quality of conditional and unconditional samples boosted by adversarial learning. Unlike previous hybrids of autoencoders and adversarial networks, the adversarial game in our approach is set up directly between the encoder and the generator, and no external mappings are trained in the process of learning. The game objective compares the divergences of each of the real and the generated data distributions with the prior distribution in the latent space. We show that direct generator-vs-encoder game leads to a tight coupling of the two components, resulting in samples and reconstructions of a comparable quality to some recently-proposed more complex architectures.

We propose a novel method to directly learn a stochastic transition operator whose repeated application provides generated samples. Traditional undirected graphical models approach this problem indirectly by learning a Markov chain model whose stationary distribution obeys detailed balance with respect to a parameterized energy function. The energy function is then modified so the model and data distributions match, with no guarantee on the number of steps required for the Markov chain to converge. Moreover, the detailed balance condition is highly restrictive: energy based models corresponding to neural networks must have symmetric weights, unlike biological neural circuits. In contrast, we develop a method for directly learning arbitrarily parameterized transition operators capable of expressing non-equilibrium stationary distributions that violate detailed balance, thereby enabling us to learn more biologically plausible asymmetric neural networks and more general non-energy based dynamical systems. The proposed training objective, which we derive via principled variational methods, encourages the transition operator to "walk back" in multi-step trajectories that start at data-points, as quickly as possible back to the original data points. We present a series of experimental results illustrating the soundness of the proposed approach, Variational Walkback (VW), on the MNIST, CIFAR-10, SVHN and CelebA datasets, demonstrating superior samples compared to earlier attempts to learn a transition operator. We also show that although each rapid training trajectory is limited to a finite but variable number of steps, our transition operator continues to generate good samples well past the length of such trajectories, thereby demonstrating the match of its non-equilibrium stationary distribution to the data distribution. Source Code: http://github.com/anirudh9119/walkback_nips17

Deep Neural Networks (DNNs) are universal function approximators providing state-of- the-art solutions on wide range of applications. Common perceptual tasks such as speech recognition, image classification, and object tracking are now commonly tackled via DNNs. Some fundamental problems remain: (1) the lack of a mathematical framework providing an explicit and interpretable input-output formula for any topology, (2) quantification of DNNs stability regarding adversarial examples (i.e. modified inputs fooling DNN predictions whilst undetectable to humans), (3) absence of generalization guarantees and controllable behaviors for ambiguous patterns, (4) leverage unlabeled data to apply DNNs to domains where expert labeling is scarce as in the medical field. Answering those points would provide theoretical perspectives for further developments based on a common ground. Furthermore, DNNs are now deployed in tremendous societal applications, pushing the need to fill this theoretical gap to ensure control, reliability, and interpretability.