Meta tags:
description=;

Headings (most frequently used words):

netadaptv2, neural, architecture, network, abstract, downloads, bibtex, efficient, search, with, fast, super, training, and, optimization, netadapt, platform, aware, adaptation, for, mobile, applications, netadaptv1,

Text of the page (most frequently used words):
the (28), and (23), netadapt (10), #network (10), search (10), google (8), metrics (8), with (8), neural (7), for (7), mobile (7), training (7), accuracy (6), time (6), dnns (6), that (5), latency (5), netadaptv2 (5), architecture (5), discovered (5), mit (4), yang (4), tien (4), sze (4), vivienne (4), platform (4), performance (4), imagenet (4), algorithm (4), while (4), super (4), non (4), differentiable (4), adaptation (3), paper (3), cpu (3), fast (3), resource (3), this (3), direct (3), better (3), optimization (3), nas (3), steps (3), 2021 (2), inproceedings (2), author (2), howard (2), andrew (2), chen (2), zhang (2), xiao (2), alec (2), sandler (2), mark (2), adam (2), hartwig (2), title (2), aware (2), applications (2), booktitle (2), conference (2), computer (2), vision (2), month (2), year (2), bibtex (2), code (2), downloads (2), validation (2), set (2), pixel (2), flow (2), automatically (2), pre (2), trained (2), deep (2), budget (2), existing (2), algorithms (2), networks (2), not (2), reduce (2), into (2), these (2), achieves (2), trade (2), offs (2), state (2), art (2), higher (2), abstract (2), liao (2), lun (2), efficient (2), three (2), innovations (2), spent (2), each (2), step (2), supporting (2), propose (2), depth (2), layer (2), single (2), evaluating (2), sampled (2), multiple (2), total (2), dnn (2), support (2), reduction (2), limits (2), eems, eccv_2018_yang_netadapt, european, eccv, september, 2018, mobilenet, lookup, tables, estimation, illustration, work, proposes, called, adapts, given, many, simplify, based, number, macs, weights, optimizing, those, indirect, may, necessarily, such, energy, consumption, solve, problem, incorporates, its, are, evaluated, using, empirical, measurements, detailed, knowledge, toolchain, required, progressively, simplifies, until, met, maximizing, experiment, results, show, versus, both, gpu, compared, automated, simplification, image, classification, dataset, speedup, measured, inference, equal, mobilenets, cvpr_2021_yang_netadaptv2, pattern, recognition, cvpr, june, present, balance, first, channel, level, bypass, connections, merge, width, dimension, second, ordered, dropout, proposed, train, forward, backward, pass, decrease, third, multi, coordinate, descent, optimizer, considers, interplay, layers, iteration, improve, reduces, nyu, respectively, discovers, mac, than, works, moreover, outperforms, mobilenetv3, top, same, typically, consists, main, most, efforts, speed, some, cost, significant, slowdown, other, sacrificing, unbalanced, per, inability, netadaptv1, project,

Text of the page (random words):
netadapt project netadaptv1 netadaptv2 netadaptv2 efficient neural architecture search with fast super network training and architecture optimization tien ju yang yi lun liao vivienne sze mit abstract neural architecture search nas typically consists of three main steps training a super network training and evaluating sampled deep neural networks dnns and training the discovered dnn most of the existing efforts speed up some steps at the cost of a significant slowdown of other steps or sacrificing the support of non differentiable search metrics the unbalanced reduction in the time spent per step limits the total search time reduction and the inability to support non differentiable search metrics limits the performance of discovered dnns in this paper we present netadaptv2 with three innovations to better balance the time spent for each step while supporting non differentiable search metrics first we propose channel level bypass connections that merge network depth and layer width into a single search dimension to reduce the time for training and evaluating sampled dnns second ordered dropout is proposed to train multiple dnns in a single forward backward pass to decrease the time for training a super network third we propose the multi layer coordinate descent optimizer that considers the interplay of multiple layers in each iteration of optimization to improve the performance of discovered dnns while supporting non differentiable search metrics with these innovations netadaptv2 reduces the total search time by up to 5 8x on imagenet and 2 4x on nyu depth v2 respectively and discovers dnns with better accuracy latency accuracy mac trade offs than state of the art nas works moreover the discovered dnn outperforms nas discovered mobilenetv3 by 1 8 higher top 1 accuracy with the same latency algorithm flow performance on imagenet validation set mobile cpu on google pixel 1 downloads paper code bibtex inproceedings cvpr_2021_yang_netadaptv2 author yang tien ju and liao yi lun and sze vivienne title netadaptv2 efficient neural architecture search with fast super network training and architecture optimization booktitle conference on computer vision and pattern recognition cvpr month june year 2021 netadapt platform aware neural network adaptation for mobile applications tien ju yang mit andrew howard google bo chen google xiao zhang google alec go google mark sandler google vivienne sze mit hartwig adam google abstract this work proposes an algorithm called netadapt that automatically adapts a pre trained deep neural network to a mobile platform given a resource budget while many existing algorithms simplify networks based on the number of macs or weights optimizing those indirect metrics may not necessarily reduce the direct metrics such as latency and energy consumption to solve this problem netadapt incorporates direct metrics into its adaptation algorithm these direct metrics are evaluated using empirical measurements so that detailed knowledge of the platform and toolchain is not required netadapt automatically and progressively simplifies a pre trained network until the resource budget is met while maximizing the accuracy experiment results show that netadapt achieves better accuracy versus latency trade offs on both mobile cpu and mobile gpu compared with the state of the art automated network simplification algorithms for image classification on the imagenet dataset netadapt achieves up to a 1 7x speedup in measured inference latency with equal or higher accuracy on mobilenets v1 v2 illustration of netadapt algorithm flow of netadapt lookup tables for fast resource estimation performance on imagenet validation set mobilenet v1 mobile cpu on google pixel 1 downloads paper code bibtex inproceedings eccv_2018_yang_netadapt author yang tien ju and howard andrew and chen bo and zhang xiao and go alec and sandler mark and sze vivienne and adam hartwig title netadapt platform aware neural network adaptation for mobile applications booktitle the european conference on computer vision eccv month september year 2018 mit eems 2021