3,200 research outputs found
Learning the group structure of deep neural networks with an expectation maximization method
No full textMany recent deep learning research work use very deep neural networks exploiting huge amount of parameters. It results in the strong expressive power, however, it also brings issues such as overfitting to training data, increasing memory burden and requiring excessive computations. In this paper, we propose an expectation maximization method to learn the group structure of deep neural networks with a group regularization principle to resolve those issues. Our method clusters the neurons in a layer based on how they are connected to the neurons in the next layer using a mixture model and the neurons in the next layer based on which group in the current layer they are most strongly connected to. Our expectation maximization method uses the Gaussian mixture model to keep the most salient connections and remove others to acquire a grouped weight matrix in a block diagonal matrix form. We refine our method further to cluster the kernels of convolutional neural networks (CNNs). We define the representative value of each kernel and build a representative matrix. The matrix is then grouped and the kernels are pruned out based on the group structure of the representative matrix. In experiments, we applied our method to fully-connected networks, 1-dimensional CNNs, and 2-dimensional CNNs and compared with baseline deep neural networks in MNIST, CIFAR-10, and United States groundwater datasets with respect to the number of parameters and classification and regression accuracy. We show that our method can reduce the number of parameters significantly without loss of accuracy and outperform the baseline models
Marangoni‐Driven Patterning in Polymer Thin Film Supported on Shrinking Substrate for Resolution Enhancement
No full textThe Marangoni effect causes liquids to flow toward localized regions with higher surface tension. In a polymer thin film, the flow induced by photochemically programmed surface tension gradients can be harnessed to manufacture patterned surfaces. Patterned polymer films are particularly useful for controlling adhesion, improving photonic device efficiency, and directing cellular alignment. In general, a broader range of accessible pattern resolutions and/or aspect ratios ensures a broader range of applications. However, because of the process flow for pattern formation, the final pattern periodicity of the Marangoni-driven features matches that of the initially prescribed surface energy pattern. To achieve better resolution without using sophisticated and complex tools, a shrinking (or pre-strained) polymer film was used as a substrate. The shrinking substrate could “contract or densify” the features along the substrate plane. Consequently, the resolution of the patterns formed on the shrinking substrate was improved by the shrinkage rate of the substrate compared with those formed on the non-shrinking substrate (i.e., silicon wafer).
Franny Choi, 41st Annual ODU Literary Festival
No full textFranny Choi is a queer, Korean-American poet, playwright, teacher, organizer, pottymouth, GryffinClaw, and general overachiever. She is the author of Floating, Brilliant, Gone (2014), and a chapbook, Death by Sex Machine (2017). She has received awards from the Poetry Foundation and the Helen Zell Writers Program, as well as fellowships from the Vermont Studio Center and the Rhode Island State Council on the Arts. Her poems have appeared in journals including Poetry magazine, American Poetry Review, New England Review, and her work has been featured by the Huffington Post, PBS NewsHour, and Angry Asian Man
Injectable tissue prosthesis for instantaneous closed-loop rehabilitation
No full textTo construct tissue-like prosthetic materials, soft electroactive hydrogels are the best candidate owing to their physiological mechanical modulus, low electrical resistance and bidirectional stimulating and recording capability of electrophysiological signals from biological tissues1,2. Nevertheless, until now, bioelectronic devices for such prostheses have been patch type, which cannot be applied onto rough, narrow or deep tissue surfaces3–5. Here we present an injectable tissue prosthesis with instantaneous bidirectional electrical conduction in the neuromuscular system. The soft and injectable prosthesis is composed of a biocompatible hydrogel with unique phenylborate-mediated multiple crosslinking, such as irreversible yet freely rearrangeable biphenyl bonds and reversible coordinate bonds with conductive gold nanoparticles formed in situ by cross-coupling. Closed-loop robot-assisted rehabilitation by injecting this prosthetic material is successfully demonstrated in the early stage of severe muscle injury in rats, and accelerated tissue repair is achieved in the later stage. © 2023, The Author(s), under exclusive licence to Springer Nature Limited.11Nsciescopu
Signal and Power Integrity (SI/PI) Analysis of Heterogeneous Integration Using Embedded Multi-die Interconnect Bridge (EMIB) Technology for High Bandwidth Memory (HBM)
No full textPower Distribution Network (PDN) Design and Analysis of A Single and. Double-Sided High Bandwidth Memory (HBM) Interposer for 2.5D Terabtye/s Bandwidth System
No full textInjection‐on‐Skin Granular Adhesive for Interactive Human–Machine Interface
No full textRealization of interactive human–machine interfaces (iHMI) is improved with development of soft tissue-like strain sensors beyond hard robotic exosuits, potentially allowing cognitive behavior therapy and physical rehabilitation for patients with brain disorders. Here, this study reports on a strain-sensitive granular adhesive inspired by the core–shell architectures of natural basil seeds for iHMI as well as human–metaverse interfacing. The granular adhesive sensor consists of easily fragmented hydropellets as a core and tissue-adhesive catecholamine layers as a shell, satisfying great on-skin injectability, ionic-electrical conductivity, and sensitive resistance changes through reversible yet robust cohesion among the hydropellets. Particularly, it is found that the ionic-electrical self-doping of the catecholamine shell on hydrosurfaces leads to a compact ion density of the materials. Based on these physical and electrical properties of the sensor, it is demonstrated that successful iHMI integration with a robot arm in both real and virtual environments enables robotic control by finger gesture and haptic feedback. This study expresses benefits of using granular hydrogel-based strain sensors for implementing on-skin writable bioelectronics and their bridging into the metaverse world. © 2023 Wiley-VCH GmbH.11Nsciescopu
Bias-dependent power distribution network impedance analysis with MOS capacitor
No full textBias-dependent power distribution network (PDN) impedance is analyzed with respect to MOS capacitor characteristics. The proposed analysis can be used to improve the PDN impedance of 2.5D and 3D ICs
QJE-STD-18-253.R2-Supplementary_Material – Supplemental material for Development and assessment of the Korean Author Recognition Test
No full textSupplemental material, QJE-STD-18-253.R2-Supplementary_Material for Development and assessment of the Korean Author Recognition Test by Hyosun Lee, Eunjin Seong, Wonil Choi and Matthew W Lowder in Quarterly Journal of Experimental Psychology</p
From the Margins to the Forefront: Tillie Olsen's Mediation as Figure and Author
No full text45 pg.Tillie Olsen's life experiences and self-identification as a working class woman provide a strong basis for analyzing her fiction as partly autobiographical. As she wrote, she developed her position as a recognized and award winning author into that of a literary mediator for socially marginalized subjects, actively working to represent certain conditions of exclusion due to social, racial, economic, and sexual factors during the 1970's and 1980's. Through analysis of her fiction and non-fiction texts, her use of modernist writing techniques, her purpose as a writer, and her impact on the literary canon, it becomes possible to see how she has altered the literary landscape and has made those who suffer exclusion visible and legible.Advisor(s): Choi, Helen . Committee Member(s): Marshik, Celia.Stony Brook University Libraries. SBU Graduate School in Department of English. Charles Taber (Dean of Graduate School)
- …
