User profiles for Han Cai
Han CaiMassachusetts Institute of Technology Verified email at mit.edu Cited by 7527 |
Path-level network transformation for efficient architecture search
We introduce a new function-preserving transformation for efficient neural architecture
search. This network transformation allows reusing previously trained networks and existing …
search. This network transformation allows reusing previously trained networks and existing …
Efficient architecture search by network transformation
Techniques for automatically designing deep neural network architectures such as reinforcement
learning based approaches have recently shown promising results. However, their …
learning based approaches have recently shown promising results. However, their …
Tinytl: Reduce memory, not parameters for efficient on-device learning
Efficient on-device learning requires a small memory footprint at training time to fit the tight
memory constraint. Existing work solves this problem by reducing the number of trainable …
memory constraint. Existing work solves this problem by reducing the number of trainable …
Apq: Joint search for network architecture, pruning and quantization policy
We present APQ, a novel design methodology for efficient deep learning deployment. Unlike
previous methods that separately optimize the neural network architecture, pruning policy, …
previous methods that separately optimize the neural network architecture, pruning policy, …
Proxylessnas: Direct neural architecture search on target task and hardware
Neural architecture search (NAS) has a great impact by automatically designing effective
neural network architectures. However, the prohibitive computational demand of conventional …
neural network architectures. However, the prohibitive computational demand of conventional …
Once-for-all: Train one network and specialize it for efficient deployment
We address the challenging problem of efficient inference across many devices and resource
constraints, especially on edge devices. Conventional approaches either manually design …
constraints, especially on edge devices. Conventional approaches either manually design …
Product-based neural networks for user response prediction
Predicting user responses, such as clicks and conversions, is of great importance and has
found its usage inmany Web applications including recommender systems, websearch and …
found its usage inmany Web applications including recommender systems, websearch and …
Hat: Hardware-aware transformers for efficient natural language processing
Transformers are ubiquitous in Natural Language Processing (NLP) tasks, but they are
difficult to be deployed on hardware due to the intensive computation. To enable low-latency …
difficult to be deployed on hardware due to the intensive computation. To enable low-latency …
Long text generation via adversarial training with leaked information
Automatically generating coherent and semantically meaningful text has many applications
in machine translation, dialogue systems, image captioning, etc. Recently, by combining with …
in machine translation, dialogue systems, image captioning, etc. Recently, by combining with …
Lite pose: Efficient architecture design for 2d human pose estimation
Pose estimation plays a critical role in human-centered vision applications. However, it is
difficult to deploy state-of-the-art HRNet-based pose estimation models on resource-…
difficult to deploy state-of-the-art HRNet-based pose estimation models on resource-…