Filters








3,079 Hits in 4.2 sec

Regression Planning Networks [article]

Danfei Xu, Roberto Martín-Martín, De-An Huang, Yuke Zhu, Silvio Savarese, Li Fei-Fei
2019 arXiv   pre-print
a sequence of intermediate goals that reaches the current observation.  ...  In this work, we combine the benefits of these two paradigms and propose a learning-to-plan method that can directly generate a long-term symbolic plan conditioned on high-dimensional observations.  ...  ., picking up a key that's already been used. Max Step is that the maximum of steps that the environment allows is reached.  ... 
arXiv:1909.13072v1 fatcat:rddzkduq6reptnzngp5jbiyqpq

FusionStitching: Boosting Execution Efficiency of Memory Intensive Computations for DL Workloads [article]

Guoping Long, Jun Yang, Wei Lin
2019 arXiv   pre-print
Yet we show in this work, that the performance of memory intensive computations is vital to E2E performance in practical DL models.  ...  Experimental results on six benchmarks and four industry scale practical models are encouraging.  ...  In order to avoid measuring M(V ) for each V , we use an memory bandwidth utilization model collected offline on the same hardware. We thus extrapolate M(V ) using the model, as shown in Figure 4 .  ... 
arXiv:1911.11576v1 fatcat:2rh6jowydrbtnnw3vtdgxweo4e

Disentangling Features in 3D Face Shapes for Joint Face Reconstruction and Recognition

Feng Liu, Ronghang Zhu, Dan Zeng, Qijun Zhao, Xiaoming Liu
2018 2018 IEEE/CVF Conference on Computer Vision and Pattern Recognition  
., non-identity) components in 3D face shapes explicitly and separately based on a composite 3D face shape model with latent representations.  ...  To construct training data we develop a method for fitting 3D morphable model (3DMM) to multiple 2D images of a subject.  ...  Our composite 3D shape model enables us to generate two types of 3D shapes. expression) components in 3D face shapes.  ... 
doi:10.1109/cvpr.2018.00547 dblp:conf/cvpr/0013ZZZ018 fatcat:4tq3ubeg6fd2fd5n3s3ucppwey

MOLI: Smart Conversation Agent for Mobile Customer Service

Guoguang Zhao, Jianyu Zhao, Yang Li, Christoph Alt, Robert Schwarzenberg, Leonhard Hennig, Stefan Schaffer, Sven Schmeier, Changjian Hu, Feiyu Xu
2019 Information  
Our approach combines models for question type and intent category classification with slot filling and a back-end knowledge base for filtering and ranking answers, and uses a dialog framework to actively  ...  Developing a dialogue system in this domain is challenging due to the broad variety of user questions.  ...  In this model, each w i ∈ t k was represented by an embedding e i ∈ R d that we obtain from pre-trained distributed word representations E = [e 1 , . . . , e W ].  ... 
doi:10.3390/info10020063 fatcat:3fxcyfqzyrfshl7unso7zwjuk4

Deep reinforcement learning with relational inductive biases

Vinícius Flores Zambaldi, David Raposo, Adam Santoro, Victor Bapst, Yujia Li, Igor Babuschkin, Karl Tuyls, David P. Reichert, Timothy P. Lillicrap, Edward Lockhart, Murray Shanahan, Victoria Langston (+4 others)
2019 International Conference on Learning Representations  
during training.  ...  We introduce an approach for augmenting model-free deep reinforcement learning agents with a mechanism for relational reasoning over structured representations, which improves performance, learning efficiency  ...  ., unlocks(key, lock) -then this function should be able to generalize to key-lock combinations that it has never observed during training.  ... 
dblp:conf/iclr/ZambaldiRSBLBTR19 fatcat:wxgbywr5mfbflcz5dmlynjpvp4

Generative Adversarial Transformers [article]

Drew A. Hudson, C. Lawrence Zitnick
2022 arXiv   pre-print
of compositional representations of objects and scenes.  ...  We introduce the GANformer, a novel and efficient type of transformer, and explore it for the task of visual generative modeling.  ...  We use a pre-trained object detector on generated CLEVR scenes to extract the objects and properties within each sample.  ... 
arXiv:2103.01209v4 fatcat:g567mweiwbb4heondushwjzp5m

Disentangling Features in 3D Face Shapes for Joint Face Reconstruction and Recognition [article]

Feng Liu, Ronghang Zhu, Dan Zeng, Qijun Zhao, Xiaoming Liu
2018 arXiv   pre-print
., non-identity) components in 3D face shapes explicitly and separately based on a composite 3D face shape model with latent representations.  ...  To construct training data we develop a method for fitting 3D morphable model (3DMM) to multiple 2D images of a subject.  ...  Our composite 3D shape model enables us to generate two types of 3D shapes. expression) components in 3D face shapes.  ... 
arXiv:1803.11366v1 fatcat:47knonszdnhtposmrmhjpxdsfe

Biological Structure and Function Emerge from Scaling Unsupervised Learning to 250 Million Protein Sequences [article]

Alexander Rives, Siddharth Goyal, Joshua Meier, Demi Guo, Myle Ott, C. Lawrence Zitnick, Jerry Ma, Rob Fergus
2019 bioRxiv   pre-print
In the field of artificial intelligence, a combination of scale in data and model capacity enabled by unsupervised learning has led to major advances in representation learning and statistical generation  ...  To this end we use unsupervised learning to train a deep contextual language model on 86 billion amino acids across 250 million sequences spanning evolutionary diversity.  ...  ACKNOWLEDGMENTS We thank Tristan Bepler, Yilun Du, Anika Gupta, Omer Levy, Ethan Perez, Neville Sanjana, and Emily Wang for insightful comments and discussions in the course of this work.  ... 
doi:10.1101/622803 fatcat:gpcfv4go4fb4rnhu3fxiwq2nmi

CHARM: A Hierarchical Deep Learning Model for Classification of Complex Human Activities Using Motion Sensors [article]

Eric Rosen, Doruk Senkal
2022 arXiv   pre-print
In this paper, we report a hierarchical deep learning model for classification of complex human activities using motion sensors.  ...  In contrast to traditional Human Activity Recognition (HAR) models used for event-based activity recognition, such as step counting, fall detection, and gesture identification, this new deep learning model  ...  In [14] , frequent patterns from low-level actions were mined to construct intermediate representations for complex activity recognition.  ... 
arXiv:2207.07806v1 fatcat:oclcaj4stre6zj4az2espsekru

Modelling, Visualising and Summarising Documents with a Single Convolutional Neural Network [article]

Misha Denil and Alban Demiraj and Nal Kalchbrenner and Phil Blunsom and Nando de Freitas
2014 arXiv   pre-print
Capturing the compositional process which maps the meaning of words to that of documents is a central challenge for researchers in Natural Language Processing and Information Retrieval.  ...  We introduce a model that is able to represent the meaning of documents by embedding them in a low dimensional vector space, while preserving distinctions of word and sentence order crucial for capturing  ...  Word embeddings learned by neural networks also serve as an excellent general purpose representation for words that can be used in non-neural models.  ... 
arXiv:1406.3830v1 fatcat:7qlivzbg75fszmzxgmlqnwmw24

Probing Across Time: What Does RoBERTa Know and When? [article]

Leo Z. Liu, Yizhong Wang, Jungo Kasai, Hannaneh Hajishirzi, Noah A. Smith
2021 arXiv   pre-print
Reasoning abilities are, in general, not stably acquired.  ...  Models of language trained on very large corpora have been demonstrated useful for NLP.  ...  Due to the large amount of training involved, we choose 14 intermediate checkpoints.  ... 
arXiv:2104.07885v2 fatcat:42k2l4bfl5gmldtzeurevfivp4

DeRF: Decomposed Radiance Fields [article]

Daniel Rebain, Wei Jiang, Soroosh Yazdani, Ke Li, Kwang Moo Yi, Andrea Tagliasacchi
2020 arXiv   pre-print
Yet, generating these images is very computationally intensive, limiting their applicability in practical scenarios.  ...  This allows us near-constant inference time regardless of the number of decomposed parts.  ...  Figure 5 . 5 Compositing with the Painter's algorithm -Visualization of the intermediate steps of the compositing process.  ... 
arXiv:2011.12490v1 fatcat:he5mkvyvrrbsfi5txa5degzo5i

Compositional Transformers for Scene Generation [article]

Drew A. Hudson, C. Lawrence Zitnick
2021 arXiv   pre-print
We introduce the GANformer2 model, an iterative object-oriented transformer, explored for the task of generative modeling.  ...  The network incorporates strong and explicit structural priors, to reflect the compositional nature of visual scenes, and synthesizes images through a sequential process.  ...  intermediate representation.  ... 
arXiv:2111.08960v1 fatcat:mevc72ear5d77igl5hln72s6hm

Reinforcement Learning with Prototypical Representations [article]

Denis Yarats, Rob Fergus, Alessandro Lazaric, Lerrel Pinto
2021 arXiv   pre-print
We pre-train these task-agnostic representations and prototypes on environments without downstream task information.  ...  Unfortunately, in RL, representation learning is confounded with the exploratory experience of the agent -- learning a useful representation requires diverse data, while effective exploration is only possible  ...  DrQ uses task reward from the outset. Plan2Explore 2 , being model-based, uses an intermediate methodology, described in Section 5.1.  ... 
arXiv:2102.11271v2 fatcat:b4g7sqj4xrcqrdxikkpdw72hfa

Data Curation with Deep Learning [Vision] [article]

Saravanan Thirumuruganathan, Nan Tang, Mourad Ouzzani, AnHai Doan
2019 arXiv   pre-print
In most organizations, data curation plays an important role so as to fully unlock the value of big data.  ...  Meanwhile, deep learning is making strides in achieving remarkable successes in multiple areas, such as image recognition, natural language processing, and speech recognition.  ...  These pre-trained models can be used in two ways: 1. feature extraction where these are used to extract generic features that are fed to a separate classifier for the task at hand; 2. fine-tuning where  ... 
arXiv:1803.01384v2 fatcat:ymch4jazxzanzpv7dbmhl5beiy
« Previous Showing results 1 — 15 out of 3,079 results