We present 4DiM, a cascaded diffusion model for 4D novel view synthesis (NVS), supporting generation with arbitrary camera trajectories and timestamps, in natural scenes, conditioned on one or more images. With a novel architecture and sampling procedure, we enable training on a mixture of 3D (with camera pose), 4D (pose+time) and video (time but no pose) data, which greatly improves generalization to unseen images and camera pose trajectories over prior works which generally operate in limited domains (e.g., object centric). 4DiM is the first-ever NVS method with intuitive metric-scale camera pose control enabled by our novel calibration pipeline for structure-from-motion-posed data. Experiments demonstrate that 4DiM outperforms prior 3D NVS models both in terms of image fidelity and pose alignment, while also enabling the generation of scene dynamics. 4DiM provides a general framework for a variety of tasks including single-image-to-3D, two-image-to-video (interpolation and extrapolation), and pose-conditioned video-to-video translation, which we illustrate qualitatively on a variety of scenes. See https://4d-diffusion.github.io for video samples.

We present VIDIM, a generative model for video inter-polation, which creates short videos given a start and end frame. In order to achieve high fidelity and generate motions unseen in the input data, VIDIM uses cascaded diffusion models to first generate the target video at low resolution, and then generate the high-resolution video conditioned on the low-resolution generated video. We compare VIDIM to previous state-of-the-art methods on video interpolation, and demonstrate how such works fail in most settings where the underlying motion is complex, nonlinear, or ambiguous while VIDIM can easily handle such cases. We additionally demonstrate how classifier-free guidance on the start and end frame and conditioning the super-resolution model on the original high-resolution frames without additional parameters unlocks high-fidelity results. VIDIM is fast to sample from as it jointly denoises all the frames to be generated, requires less than a billion parameters per diffusion model to produce compelling results, and still enjoys scalability and improved quality at larger parameter counts. Please see our project page at vidim-interpolation.github.io.

3D reconstruction methods such as Neural Radiance Fields (NeRFs) excel at rendering photorealistic novel views of complex scenes. However, recovering a high-quality NeRF typically requires tens to hundreds of input images, resulting in a time-consuming capture process. We present ReconFusion to reconstruct real-world scenes using only a few photos. Our approach leverages a diffusion prior for novel view synthesis, trained on synthetic and multiview datasets, which regularizes a NeRF-based 3D reconstruction pipeline at novel camera poses beyond those captured by the set of input images. Our method synthesizes realistic geometry and texture in underconstrained regions while preserving the appearance of observed regions. We perform an extensive evaluation across various real-world datasets, including forward-facing and 360-degree scenes, demonstrating significant performance improvements over previous few-view NeRF reconstruction approaches. Please see our project page at reconfusion.github. io.

BACKGROUND: Life science research is moving quickly towards large-scale experimental designs that are comprised of multiple tissues, time points, and samples. Omic time-series experiments offer answers to three big questions: what collective patterns do most analytes follow, which analytes follow an identical pattern or synchronize across multiple cohorts, and how do biological functions evolve over time. Existing tools fall short of robustly answering and visualizing all three questions in a unified interface. RESULTS: Functional Heatmap offers time-series data visualization through a Master Panel page, and Combined page to answer each of the three time-series questions. It dissects the complex multi-omics time-series readouts into patterned clusters with associated biological functions. It allows users to identify a cascade of functional changes over a time variable. Inversely, Functional Heatmap can compare a pattern with specific biology respond to multiple experimental conditions. All analyses are interactive, searchable, and exportable in a form of heatmap, line-chart, or text, and the results are easy to share, maintain, and reproduce on the web platform. CONCLUSIONS: Functional Heatmap is an automated and interactive tool that enables pattern recognition in time-series multi-omics assays. It significantly reduces the manual labour of pattern discovery and comparison by transferring statistical models into visual clues. The new pattern recognition feature will help researchers identify hidden trends driven by functional changes using multi-tissues/conditions on a time-series fashion from omic assays.

Text normalization is an important enabling technology for several NLP tasks.Recently, neural-network-based approaches have outperformed well-established models in this task.However, in languages other than English, there has been little exploration in this direction.Both the scarcity of annotated data and the complexity of the language increase the difficulty of the problem.To address these challenges, we use a sequence-to-sequence model with character-based attention, which in addition to its self-learned character embeddings, uses word embeddings pre-trained with an approach that also models subword information.This provides the neural model with access to more linguistic information especially suitable for text normalization, without large parallel corpora.We show that providing the model with word-level features bridges the gap for the neural network approach to achieve a state-of-the-art F 1 score on a standard Arabic language correction shared task dataset.

BACKGROUND: Network medicine aims to map molecular perturbations of any given diseases onto complex networks with functional interdependencies that underlie a pathological phenotype. Furthermore, investigating the time dimension of disease progression from a network perspective is key to gaining key insights to the disease process and to identify diagnostic or therapeutic targets. Existing platforms are ineffective to modularize the large complex systems into subgroups and consolidate heterogeneous data to web-based interactive animation. RESULTS: We have developed PanoromiX platform, a data-agnostic dynamic interactive visualization web application, enables the visualization of outputs from genome based molecular assays onto modular and interactive networks that are correlated with any pathophenotypic data (MRI, Xray, behavioral, etc.) over a time course all in one pane. As a result, PanoromiX reveals the complex organizing principles that orchestrate a disease-pathology from a gene regulatory network (nodes, edges, hubs, etc.) perspective instead of snap shots of assays. Without extensive programming experience, users can design, share, and interpret their dynamic networks through the PanoromiX platform with rich built-in functionalities. CONCLUSIONS: This emergent tool of network medicine is the first to visualize the interconnectedness of tailored genome assays to pathological networks and phenotypes for cells or organisms in a data-agnostic manner. As an advanced network medicine tool, PanoromiX allows monitoring of panel of biomarker perturbations over the progression of diseases, disease classification based on changing network modules that corresponds to specific patho-phenotype as opposed to clinical symptoms, systematic exploration of complex molecular interactions and distinct disease states via regulatory network changes, and the discovery of novel diagnostic and therapeutic targets.

This paper presents the network bandwidth utilization results of a hybrid peer-to-peer massively multiplayer framework. Previous work in this area has focused on model proposals and simulation based results. The work in this paper is novel in that it also describes a real-world implementation and presents results from actual network traces. The implementation demonstrates the potential for the model and the results show the network bandwidth utilization is well within the current capabilities of residential broadband.

UNLABELLED: We built a novel web-based platform for performing discrete molecular dynamics simulations of proteins. In silico protein folding involves searching for minimal frustration in the vast conformational landscape. Conventional approaches for simulating protein folding insufficiently address the problem of simulations in relevant time and length scales necessary for a mechanistic understanding of underlying biomolecular phenomena. Discrete molecular dynamics (DMD) offers an opportunity to bridge the size and timescale gaps and uncover the structural and biological properties of experimentally undetectable protein dynamics. The iFold server supports large-scale simulations of protein folding, thermal denaturation, thermodynamic scan, simulated annealing and p(fold) analysis using DMD and coarse-grained protein model with structure-based Gō-interactions between amino acids. AVAILABILITY: http://ifold.dokhlab.org