E LASTIC-I NFOGAN: U NSUPERVISEDD ISENTANGLED R EPRESENTATIONL EARNING IN I MBALANCEDD ATA Utkarsh Ojha 1, Krishna Kumar Singh 1, Cho-Jui Hsieh 2, and Y ong Jae Lee 1 1 University of California, Davis 2 University of California, Los Angeles A BSTRACT We propose a novel unsupervised generative model, Elastic-InfoGAN, that learns to disentangle object identity from other low-level aspects in class-imbalanced datasets. We first investigate the issues surrounding the assumptions about uniformity made by InfoGAN (Chen et al. (2016)), and demonstrate its ineffectiveness to properly disentangle object identity in imbalanced data. Our key idea is to make the discovery of the discrete latent factor of variation invariant to identity-preserving transformations in real images, and use that as the signal to learn the latent distribution's parameters. Experiments on both artificial (MNIST) and real-world (Y ouTube-Faces) datasets demonstrate the effectiveness of our approach in imbalanced data by: (i) better disentanglement of object identity as a latent factor of variation; and (ii) better approximation of class imbalance in the data, as reflected in the learned parameters of the latent distribution. Recent deep neural network based models such as Generative Adversarial Networks (Goodfellow et al. (2014); Salimans et al. (2016); Radford et al. (2016)) and V ariational Autoen-coders (Kingma & Welling (2014); Higgins et al. (2017)) have led to promising results in generating realistic samples for high-dimensional and complex data such as images. More advanced models show how to discover disentangled representations (Y an et al. (2016); Chen et al. (2016); Tran et al. (2017); Hu et al. (2018); Singh et al. (2019)), in which different latent dimensions can be made to represent independent factors of variation (e.g., pose, identity) in the data (e.g., human faces). InfoGAN (Chen et al. (2016)) in particular, tries to learn an unsupervised disentangled representation by maximizing the mutual information between the discrete or continuous latent variables and the corresponding generated samples. For discrete latent factors (e.g., digit identities), it assumes that they are uniformly distributed in the data, and approximates them accordingly using a fixed uniform categorical distribution. Although this assumption holds true for many existing benchmark datasets (e.g., MNIST LeCun (1998)), real-word data often follows a long-tailed distribution and rarely exhibits perfect balance between the categories. Indeed, applying InfoGAN on imbalanced data can result in incoherent groupings, since it is forced to discover potentially nonexistent factors that are uniformly distributed in the data; see Figure 1.

We study the problem of learning and optimizing through physical simulations via differentiable programming. We present DiffTaichi, a new differentiable programming language tailored for building high-performance differentiable physical simulations. We demonstrate the performance and productivity of our language in gradient-based learning and optimization tasks on 10 different physical simulators. For example, a differentiable elastic object simulator written in our language is 4.2x faster than the hand-engineered CUDA version yet runs as fast, and is 188x faster than TensorFlow. Using our differentiable programs, neural network controllers are typically optimized within only tens of iterations. Finally, we share the lessons learned from our experience developing these simulators, that is, differentiating physical simulators does not always yield useful gradients of the physical system being simulated. We systematically study the underlying reasons and propose solutions to improve gradient quality.

This paper introduces Selective-Backprop, a technique that accelerates the training of deep neural networks (DNNs) by prioritizing examples with high loss at each iteration. Selective-Backprop uses the output of a training example's forward pass to decide whether to use that example to compute gradients and update parameters, or to skip immediately to the next example. By reducing the number of computationally-expensive backpropagation steps performed, Selective-Backprop accelerates training. Evaluation on CIFAR10, CIFAR100, and SVHN, across a variety of modern image models, shows that Selective-Backprop converges to target error rates up to 3.5x faster than with standard SGD and between 1.02--1.8x faster than a state-of-the-art importance sampling approach. Further acceleration of 26% can be achieved by using stale forward pass results for selection, thus also skipping forward passes of low priority examples.

We propose a new family of efficient and expressive deep generative models of graphs, called Graph Recurrent Attention Networks (GRANs). Our model generates graphs one block of nodes and associated edges at a time. The block size and sampling stride allow us to trade off sample quality for efficiency. Compared to previous RNN-based graph generative models, our framework better captures the auto-regressive conditioning between the already-generated and to-be-generated parts of the graph using Graph Neural Networks (GNNs) with attention. This not only reduces the dependency on node ordering but also bypasses the long-term bottleneck caused by the sequential nature of RNNs. Moreover, we parameterize the output distribution per block using a mixture of Bernoulli, which captures the correlations among generated edges within the block. Finally, we propose to handle node orderings in generation by marginalizing over a family of canonical orderings. On standard benchmarks, we achieve state-of-the-art time efficiency and sample quality compared to previous models. Additionally, we show our model is capable of generating large graphs of up to 5K nodes with good quality. To the best of our knowledge, GRAN is the first deep graph generative model that can scale to this size. Our code is released at: https://github.com/lrjconan/GRAN.

We present a proxy dataset of vital signs with class labels indicating patient transitions from the ward to intensive care units called Ward2ICU. Patient privacy is protected using a Wasserstein Generative Adversarial Network to implicitly learn an approximation of the data distribution, allowing us to sample synthetic data. The quality of data generation is assessed directly on the binary classification task by comparing specificity and sensitivity of an LSTM classifier on proxy and original datasets. We initialize a discussion of unintentionally disclosing commercial sensitive information and propose a solution for a special case through class label balancing

The relation of individual parameters to the network's output is highly nonlinear and is generally unclear to an external observer. Consequently, it has been widely supposed in the field that it is impossible to recover the parameters of a network merely by observing its output on different inputs. Beyond informing our understanding of deep learning, going from function to parameters could have serious implications for security and privacy. In many deployed deep learning systems, the output is freely available, but the network used to generate that output is not disclosed. The ability to uncover a confidential network not only would make it available for public use but could even expose data used to train the network if such data could be reconstructed from the network's weights. This topic also has implications for the study of biological neural networks. Experimental neuroscientists can record some variables within the brain (e.g. the output of a complex cell in primary visual cortex) but not others (e.g. the pre-synaptic simple cells), and many biological neurons appear to be well modeled as the ReLU of a linear combination of their inputs (Chance et al., 2002). It would be highly useful if we could reverse engineer the internal components of a neural circuit based on recordings of the output and our choice of input stimuli. 1 arXiv:1910.00744v1

Machine learning (ML) algorithms, especially deep neural networks, have demonstrated success in several domains. However, several types of attacks have raised concerns about deploying ML in safety-critical domains, such as autonomous driving and security. An attacker perturbs a data point slightly in the concrete feature space (e.g., pixel space) and causes the ML algorithm to produce incorrect output (e.g. a perturbed stop sign is classified as a yield sign). These perturbed data points are called adversarial examples, and there are numerous algorithms in the literature for constructing adversarial examples and defending against them. In this paper we explore semantic adversarial examples (SAEs) where an attacker creates perturbations in the semantic space representing the environment that produces input for the ML model. For example, an attacker can change the background of the image to be cloudier to cause misclassification. We present an algorithm for constructing SAEs that uses recent advances in differential rendering and inverse graphics.

In this paper, we present NESTA, a specialized Neural engine that significantly accelerates the computation of convolution layers in a deep convolutional neural network, while reducing the computational energy. NESTA reformats Convolutions into $3 \times 3$ batches and uses a hierarchy of Hamming Weight Compressors to process each batch. Besides, when processing the convolution across multiple channels, NESTA, rather than computing the precise result of a convolution per channel, quickly computes an approximation of its partial sum, and a residual value such that if added to the approximate partial sum, generates the accurate output. Then, instead of immediately adding the residual, it uses (consumes) the residual when processing the next batch in the hamming weight compressors with available capacity. This mechanism shortens the critical path by avoiding the need to propagate carry signals during each round of computation and speeds up the convolution of each channel. In the last stage of computation, when the partial sum of the last channel is computed, NESTA terminates by adding the residual bits to the approximate output to generate a correct result.

Every year thousands of patients are diagnosed with a glioma, a type of malignant brain tumor. Physicians use MR images as a key tool in the diagnosis and treatment of these patients. Neural networks show great potential to aid physicians in the medical image analysis. This study investigates the use of varying amounts of synthetic brain T1-weighted (T1), post-contrast T1-weighted (T1Gd), T2-weighted (T2), and T2 Fluid Attenuated Inversion Recovery (FLAIR) MR images created by a generative adversarial network to overcome the lack of annotated medical image data in training separate 2D U-Nets to segment enhancing tumor, peritumoral edema, and necrosis (non-enhancing tumor core) regions on gliomas. These synthetic MR images were assessed quantitively (SSIM=0.79) and qualitatively by a physician who found that the synthetic images seem stronger for delineation of structural boundaries but struggle more when gradient is significant, (e.g. edema signal in T2 modalities). Multiple 2D U-Nets were trained with original BraTS data and differing subsets of a quarter, half, three-quarters, and all synthetic MR images. There was not an obvious correlation between the improvement of values of the metrics in separate validation dataset for each structure and amount of synthetic data added, there is a strong correlation between the amount of synthetic data added and the number of best overall validation metrics. In summary, this study showed ability to generate high quality synthetic Flair, T2, T1, and T1CE MR images using the GAN. Using the synthetic MR images showed encouraging results to improve the U-Net segmentation performance which has the potential to address the scarcity of readily available medical images.

State of the art Deep Neural Networks (DNN) can now achieve above human level accuracy on image classification tasks. However their outstanding performances come along with a complex inference mechanism making them arduously interpretable models. In order to understand the underlying prediction rules of DNNs, Dhamdhere et al. propose an interpretability method to break down a DNN prediction score as sum of its hidden unit contributions, in the form of a metric called conductance. Analyzing conductances of DNN hidden units, we find out there is a difference in how wrong and correct predictions are inferred. We identify distinguishable patterns of hidden unit activations for wrong and correct predictions. We then use an error detector in the form of a binary classifier on top of the DNN to automatically discriminate wrong and correct predictions of the DNN based on their hidden unit activations. Detected wrong predictions are discarded, increasing the model robustness. A different approach to distinguish wrong and correct predictions of DNNs is proposed by Wang et al. whose method is based on the premise that input samples leading a DNN into making wrong predictions are less stable to the DNN weight changes than correctly classified input samples. In our study, we compare both methods and find out by combining them that better detection of wrong predictions can be achieved.