Results 51 to 60 of about 118,488 (274)

Altitude Training: Strong Bounds for Single-Layer Dropout [PDF]

, 2014
Dropout training, originally designed for deep neural networks, has been successful on high-dimensional single-layer natural language tasks. This paper proposes a theoretical explanation for this phenomenon: we show that, under a generative Poisson topic
Fithian, William, Liang, Percy, Wager, Stefan, Wang, Sida   +3 more
core   +1 more source

Texoskeletons: Developing the Fundamental Technologies for Creating Intelligent Soft Robotic Clothing With Integrated 1D Sensors and Actuators

Advanced Functional Materials, EarlyView.
ABSTRACT Traditional wearable exoskeletons rely on rigid structures, which limit comfort, flexibility, and everyday usability. This work introduces the fundamental technologies to create the first soft, lightweight, intelligent textile‐based exoskeletons (Texoskeletons) built using 1D sensors and actuators.
Amy Lukomiak, Rameesh Bulathsinghala, Jack Varga, Shuxin Meng, Kalana Marasinghe, Imaad Refai, Biyon Fernando, Amitkumar Patel, Chamika M. Halloluwa‐Arachchige, Carlos Oliveira, Arash M. Shahidi, Zahra Rahemtulla, Alexander Turner, Ziyun Ding, Bernard X. W. Liew, Andy Kerr, Ezio Preatoni, Theo Hughes‐Riley, Ishara Dharmasena, Pasindu Lugoda   +19 more
wiley   +1 more source

Self‐Powered Flexible Triboelectric‐Gated Ion‐Gel Transistor for Neuromorphic Tactile Sensing and Human Activity Recognition

Advanced Materials, EarlyView.
A fully flexible ion‐gel‐gated graphene‐channel transistor driven by a triboelectric nanogenerator enables self‐powered tactile sensing and synaptic learning. Mimicking spike‐rate‐dependent plasticity, the device exhibits frequency‐selective potentiation and depression, supporting rate‐coded neuromorphic computation even under flex.
Hanseong Cho, Seoyeon Park, Youngmin Lee, Sejoon Lee   +3 more
wiley   +1 more source

Spectral Adaptive Dropout: Frequency-Based Regularization for Improved Generalization

Information
Deep neural networks are often susceptible to overfitting, necessitating effective regularization techniques. This paper introduces Spectral Adaptive Dropout, a novel frequency-based regularization technique that dynamically adjusts dropout rates based ...
Zhigao Huang, Musheng Chen, Shiyan Zheng
doaj   +1 more source

Curriculum Dropout

, 2017
Dropout is a very effective way of regularizing neural networks. Stochastically "dropping out" units with a certain probability discourages over-specific co-adaptations of feature detectors, preventing overfitting and improving network generalization ...
Cavazza, Jacopo, Morerio, Pietro, Murino, Vittorio, Vidal, Rene, Volpi, Riccardo   +4 more
core   +1 more source

Characteristics of Monte Carlo Dropout in Wide Neural Networks

CoRR, 2020
Accepted at the ICML 2020 workshop for Uncertainty and Robustness in Deep ...
Joachim Sicking, Maram Akila, Tim Wirtz, Sebastian Houben, Asja Fischer   +4 more
openaire   +2 more sources

A Scalable Perovskite Platform With Multi‐State Photoresponsivity for In‐Sensor Saliency Detection

Advanced Materials, EarlyView.
A scalable in‐sensor computing platform (32 × 32 array) with ultra‐low variability is developed by incorporating ferroelectric copolymers into halide perovskite thin films. These devices achieve 1000 programmable photoresponsivity states and high thermal reliability.
Xuechao Xing, Anubhab Tripathi, Si En Ng, Natalia Yantara, Yue Gong, Qinjie Wu, Yeow Boon Tay, Wen Siang Lew, Yang Chai, Arindam Basu, Nripan Mathews   +10 more
wiley   +1 more source

Flow‐Adaptive Gas Sensing Enabled Using a Uniform Au Nanosheet Sensor Array and a Neural Network Inference

Advanced Materials Technologies, EarlyView.
Integrated Au nanosheet sensor array enables simultaneous inference of gas concentration and flow rate via deep neural network analysis, without external flow control. ABSTRACT Gas sensor responses are considerably affected by gas flow rates, thereby inhibiting the accurate detection of target gas concentrations in variable‐flow applications such as ...
Taro Kato, Takahisa Tanaka, Yusuke Hamanaka, Ryo Toyoshima, Ken Uchida   +4 more
wiley   +1 more source

Dataset distillation with stochastic neural networks

Complex & Intelligent Systems
Dataset distillation aims to synthesize tiny and high-fidelity data that contains the most important information of a given target dataset. Recent studies primarily used gradient-matching based methods to attain practical performance.
Zeyuan Wang, Tianyuan Yu, Liang Bai, Jun Tang, Yanming Guo, Yirun Ruan   +5 more
doaj   +1 more source

Bayesian Dropout

, 2015
Dropout has recently emerged as a powerful and simple method for training neural networks preventing co-adaptation by stochastically omitting neurons.
Herlau, Tue, Mørup, Morten, Schmidt, Mikkel N.   +2 more
core   +1 more source