no code implementations • 4 Oct 2023 • Jared Lichtarge, Ehsan Amid, Shankar Kumar, Tien-Ju Yang, Rohan Anil, Rajiv Mathews
Federated Averaging, and many federated learning algorithm variants which build upon it, have a limitation: all clients must share the same model architecture.
no code implementations • 14 Sep 2022 • Rongmei Lin, Yonghui Xiao, Tien-Ju Yang, Ding Zhao, Li Xiong, Giovanni Motta, Françoise Beaufays
Automatic Speech Recognition models require large amount of speech data for training, and the collection of such data often leads to privacy concerns.
Automatic Speech Recognition Automatic Speech Recognition (ASR) +2
no code implementations • 6 May 2022 • Tien-Ju Yang, Yonghui Xiao, Giovanni Motta, Françoise Beaufays, Rajiv Mathews, Mingqing Chen
This paper addresses the challenges of training large neural network models under federated learning settings: high on-device memory usage and communication cost.
no code implementations • 11 Oct 2021 • Tien-Ju Yang, Dhruv Guliani, Françoise Beaufays, Giovanni Motta
This paper aims to address the major challenges of Federated Learning (FL) on edge devices: limited memory and expensive communication.
no code implementations • 7 Oct 2021 • Dhruv Guliani, Lillian Zhou, Changwan Ryu, Tien-Ju Yang, Harry Zhang, Yonghui Xiao, Francoise Beaufays, Giovanni Motta
Federated learning can be used to train machine learning models on the edge on local data that never leave devices, providing privacy by default.
Automatic Speech Recognition Automatic Speech Recognition (ASR) +2
no code implementations • CVPR 2021 • Tien-Ju Yang, Yi-Lun Liao, Vivienne Sze
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.
no code implementations • 18 Dec 2019 • Tien-Ju Yang, Vivienne Sze
This paper describes various design considerations for deep neural networks that enable them to operate efficiently and accurately on processing-in-memory accelerators.
1 code implementation • ICCV 2019 • Jyh-Jing Hwang, Stella X. Yu, Jianbo Shi, Maxwell D. Collins, Tien-Ju Yang, Xiao Zhang, Liang-Chieh Chen
The proposed SegSort further produces an interpretable result, as each choice of label can be easily understood from the retrieved nearest segments.
Ranked #10 on Unsupervised Semantic Segmentation on PASCAL VOC 2012 val (using extra training data)
1 code implementation • 8 Mar 2019 • Diana Wofk, Fangchang Ma, Tien-Ju Yang, Sertac Karaman, Vivienne Sze
In this paper, we address the problem of fast depth estimation on embedded systems.
no code implementations • 13 Feb 2019 • Tien-Ju Yang, Maxwell D. Collins, Yukun Zhu, Jyh-Jing Hwang, Ting Liu, Xiao Zhang, Vivienne Sze, George Papandreou, Liang-Chieh Chen
We present a single-shot, bottom-up approach for whole image parsing.
Ranked #32 on Panoptic Segmentation on Cityscapes val
1 code implementation • 10 Jul 2018 • Yu-Hsin Chen, Tien-Ju Yang, Joel Emer, Vivienne Sze
In this work, we present Eyeriss v2, a DNN accelerator architecture designed for running compact and sparse DNNs.
4 code implementations • ECCV 2018 • Tien-Ju Yang, Andrew Howard, Bo Chen, Xiao Zhang, Alec Go, Mark Sandler, Vivienne Sze, Hartwig Adam
This work proposes an algorithm, called NetAdapt, that automatically adapts a pre-trained deep neural network to a mobile platform given a resource budget.
3 code implementations • CVPR 2018 • Ariel Gordon, Elad Eban, Ofir Nachum, Bo Chen, Hao Wu, Tien-Ju Yang, Edward Choi
We present MorphNet, an approach to automate the design of neural network structures.
no code implementations • 27 Mar 2017 • Vivienne Sze, Yu-Hsin Chen, Tien-Ju Yang, Joel Emer
The reader will take away the following concepts from this article: understand the key design considerations for DNNs; be able to evaluate different DNN hardware implementations with benchmarks and comparison metrics; understand the trade-offs between various hardware architectures and platforms; be able to evaluate the utility of various DNN design techniques for efficient processing; and understand recent implementation trends and opportunities.
no code implementations • CVPR 2017 • Tien-Ju Yang, Yu-Hsin Chen, Vivienne Sze
With the proposed pruning method, the energy consumption of AlexNet and GoogLeNet are reduced by 3. 7x and 1. 6x, respectively, with less than 1% top-5 accuracy loss.