In this work, we demonstrate the ability to exploit sensing modalities for mitigating an unrepresented modality or for potentially re-targeting resources. This is tantamount to developing proxy sensing capabilities for multi-modal learning. In classical fusion, multiple sensors are required to capture different information about the same target. Maintaining and collecting samples from multiple sensors can be financially demanding. Additionally, the effort necessary to ensure a logical mapping between the modalities may be prohibitively limiting. We examine the scenario where we have access to all modalities during training, but only a single modality at testing. In our approach, we initialize the parameters of our single modality inference network with weights learned from the fusion of multiple modalities through both classification and GANs losses. Our experiments show that emulating a multi-modal system by perturbing a single modality with noise can help us achieve competitive results compared to using multiple modalities.

Simulating complex biological and physiological systems and predicting their behaviours under different conditions remains challenging. Breaking systems into smaller and more manageable modules can address this challenge, assisting both model development and simulation. Nevertheless, existing computational models in biology and physiology are often not modular and therefore difficult to assemble into larger models. Even when this is possible, the resulting model may not be useful due to inconsistencies either with the laws of physics or the physiological behaviour of the system. Here, we propose a general methodology for composing models, combining the energy-based bond graph approach with semantics-based annotations. This approach improves model composition and ensures that a composite model is physically plausible. As an example, we demonstrate this approach to automated model composition using a model of human arterial circulation. The major benefit is that modellers can spend more time on understanding the behaviour of complex biological and physiological systems and less time wrangling with model composition.

We propose a deep learning architecture capable of performing up to 8× single image super-resolution. Our architecture incorporates an adversarial component from the super-resolution generative adversarial networks (SRGANs) and a multi-scale learning component from the multiple scale super-resolution network (MSSRNet), which only together can recover smaller structures inherent in satellite images. To further enhance our performance, we integrate progressive growing and training to our network. This, aided by feed forwarding connections in the network to move along and enrich information from previous inputs, produces super-resolved images at scaling factors of 2, 4, and 8. To ensure and enhance the stability of GANs, we employ Wasserstein GANs (WGANs) during training. Experimentally, we find that our architecture can recover small objects in satellite images during super-resolution whereas previous methods cannot.

A learned generative model often produces biased statistics relative to the underlying data distribution. A standard technique to correct this bias is importance sampling, where samples from the model are weighted by the likelihood ratio under model and true distributions. When the likelihood ratio is unknown, it can be estimated by training a probabilistic classifier to distinguish samples from the two distributions. We employ this likelihood-free importance weighting method to correct for the bias in generative models. We find that this technique consistently improves standard goodness-of-fit metrics for evaluating the sample quality of state-of-the-art deep generative models, suggesting reduced bias. Finally, we demonstrate its utility on representative applications in a) data augmentation for classification using generative adversarial networks, and b) model-based policy evaluation using off-policy data.

Recent electron microscopy data have revealed that cardiac mitochondria are not arranged in crystalline columns but are organised with several mitochondria aggregated into columns of varying sizes spanning the cell cross-section. This raises the question-how does the mitochondrial arrangement affect the metabolite distributions within cardiomyocytes and what is its impact on force dynamics? Here, we address this question by employing finite element modeling of cardiac bioenergetics on computational meshes derived from electron microscope images. Our results indicate that heterogeneous mitochondrial distributions can lead to significant spatial variation across the cell in concentrations of inorganic phosphate, creatine (Cr) and creatine phosphate (PCr). However, our model predicts that sufficient activity of the creatine kinase (CK) system, coupled with rapid diffusion of Cr and PCr, maintains near uniform ATP and ADP ratios across the cell cross sections. This homogenous distribution of ATP and ADP should also evenly distribute force production and twitch duration with contraction. These results suggest that the PCr shuttle and associated enzymatic reactions act to maintain uniform force dynamics in the cell despite the heterogeneous mitochondrial organization. However, our model also predicts that under hypoxia activity of mitochondrial CK enzymes and diffusion of high-energy phosphate compounds may be insufficient to sustain uniform ATP/ADP distribution and hence force generation.

A Semantic Compositional Network (SCN) is developed for image captioning, in which semantic concepts (i.e., tags) are detected from the image, and the probability of each tag is used to compose the parameters in a long short-term memory (LSTM) network. The SCN extends each weight matrix of the LSTM to an ensemble of tag-dependent weight matrices. The degree to which each member of the ensemble is used to generate an image caption is tied to the image-dependent probability of the corresponding tag. In addition to captioning images, we also extend the SCN to generate captions for video clips. We qualitatively analyze semantic composition in SCNs, and quantitatively evaluate the algorithm on three benchmark datasets: COCO, Flickr30k, and Youtube2Text. Experimental results show that the proposed method significantly outperforms prior state-of-the-art approaches, across multiple evaluation metrics.

Stochastic Dual Coordinate Ascent (SDCA) has recently emerged as a state-of-the-art method for solving large-scale supervised learning problems formulated as minimization of convex loss functions. It performs iterative, random-coordinate updates to maximize the dual objective. Due to the sequential nature of the iterations, it is typically implemented as a single-threaded algorithm limited to in-memory datasets. In this paper, we introduce an asynchronous parallel version of the algorithm, analyze its convergence properties, and propose a solution for primal-dual synchronization required to achieve convergence in practice. In addition, we describe a method for scaling the algorithm to out-of-memory datasets via multi-threaded deserialization of block-compressed data. This approach yields sufficient pseudo-randomness to provide the same convergence rate as random-order in-memory access. Empirical evaluation demonstrates the efficiency of the proposed methods and their ability to fully utilize computational resources and scale to out-of-memory datasets.