Coupled Time-Domain Hydro-Elastic Simulation for Submerged Floating Tunnel Under Wave Excitations

We develop a fully-coupled time-domain hydro-elasticity model for the Submerged Floating Tunnel (SFT) based on the Discrete-Module-Beam (DMB) method. Frequency-domain simulation based on 3D potential theory results in multibody’s hydrodynamic coefficients and excitation forces for tunnel sections. Subsequently, we build the time-domain model with the multibody Cummins equation and external stiffness matrix from the Euler-Bernoulli and Saint-Venant torsion theories. We establish the mooring line model with rod theory and couple components with translational springs at their respective connection locations. We then compare the dynamic motions, wave forces, and mooring tensions between the present and Morison-equation-based elastic models under regular wave excitations at different submergence depths. The present model is especially important for the shallowly submerged tunnel in which the Morison model shows exaggerated motions, especially at high-frequency range.

Learning to Discretely Compose Reasoning Module Networks for Video Captioning

Generating natural language descriptions for videos, i.e., video captioning, essentially requires step-by-step reasoning along the generation process. For example, to generate the sentence “a man is shooting a basketball”, we need to first locate and describe the subject “man”, next reason out the man is “shooting”, then describe the object “basketball” of shooting. However, existing visual reasoning methods designed for visual question answering are not appropriate to video captioning, for it requires more complex visual reasoning on videos over both space and time, and dynamic module composition along the generation process. In this paper, we propose a novel visual reasoning approach for video captioning, named Reasoning Module Networks (RMN), to equip the existing encoder-decoder framework with the above reasoning capacity. Specifically, our RMN employs 1) three sophisticated spatio-temporal reasoning modules, and 2) a dynamic and discrete module selector trained by a linguistic loss with a Gumbel approximation. Extensive experiments on MSVD and MSR-VTT datasets demonstrate the proposed RMN outperforms the state-of-the-art methods while providing an explicit and explainable generation process. Our code is available at https://github.com/tgc1997/RMN.

D4-Modules

A module [Formula: see text] is called a [Formula: see text]-module if, whenever [Formula: see text] and [Formula: see text] are submodules of [Formula: see text] with [Formula: see text] and [Formula: see text] is a homomorphism with [Formula: see text], we have [Formula: see text]. The class of [Formula: see text]-modules contains the [Formula: see text]-modules as well as the dual-square-free (DSF) modules. Furthermore, a [Formula: see text]-module [Formula: see text] is called pseudo-discrete if [Formula: see text] is also a lifting module. In this paper, we study the [Formula: see text]-, the DSF, and the pseudo-discrete modules, and show that a pseudo-discrete module is clean iff it has the finite exchange property iff it has the full exchange property.

The Ersatz Brain Project

The Ersatz Brain Project develops programming techniques and software applications for a brain-like computing system. Its brain-like hardware architecture design is based on a select set of ideas taken from the anatomy of mammalian neo-cortex. In common with other such attempts it is based on a massively parallel, two-dimensional array of CPUs and their associated memory. The design used in this project: 1) Uses an approximation to cortical computation called the network of networks which holds that the basic computing unit in the cortex is not a single neuron but groups of neurons working together in attractor networks; 2) Assumes connections and data representations in cortex are sparse; 3) Makes extensive use of local lateral connections and topographic data representations, and 4) Scales in a natural way from small groups of neurons to the entire cortical regions. The resulting system computes effectively using techniques such as local data movement, sparse data representation, sparse connectivity, temporal coincidence, and the formation of discrete “module assemblies.” The authors discuss recent neuroscience in relation to their physiological assumptions and a set of experiments displaying what appear to be “concept-like” ensemble based cells in human cortex.