DEEPREINFORCEMENT LEARNING IN PARAMETERIZED ACTION SPACE
dqn在足球游戏中的研究
代码:Complete source code for our agent isavailable at https://github.com/mhauskn/dqn-hfo and for the RoboCup 2D domainat https://github.com/mhauskn/HFO/.
DEEPMULTIMODAL SEMANTIC EMBEDDINGS FOR SPEECH AND IMAGES
HeterogeneousKnowledge Transfer in Video Emotion Recognition, Attribution and Summarization
WordEmbedding based Correlation Model for Question/Answer Matching
improvethe matching (word pair) accuracy
FeatureLearning based Deep Supervised Hashing with Pairwise Labels
1. 网络更深了
2. 前向fc8用了sign
3. 后向用了relex sign
DeepMetric Learning via Lifted Structured Feature Embedding
从pair wised-> triplet->(list, lifted structured)
cnn与metric learning更多更深入的结合
INTEGRATINGDEEP FEATURES FOR MATERIAL RECOGNITION
比较实用的工程技巧
WIDERFACE: A Face Detection Benchmark
TheWIDER FACE dataset consists of 393, 703 labeled face bounding boxes in 32, 203images
新的人脸数据集
Single-viewto Multi-view: Reconstructing Unseen Views with a Convolutional Network
可以在更多需要重构的场景中去尝试
TRAININGCNNS WITH LOW-RANK FILTERS FOR EFFICIENT IMAGE CLASSIFICATION
带效果损失的模型大小与前向时间优化
COMPRESSIONOF DEEP CONVOLUTIONAL NEURAL NETWORKS FOR FAST AND LOW POWER MOBILEAPPLICATIONS
模型运行时与大小优化 for 智能设备
ImagesDon’t Lie: Transferring Deep Visual Semantic Features to Large-Scale MultimodalLearning to Rank
Multimodal Listing Embedding+tuple方式Learning To Rank
目前看到的电商里的第一篇multimodal

本文探讨了深度强化学习在参数化行动空间中的应用,具体以足球游戏中的DQN算法为例进行深入研究,并涉及多模态语义嵌入、情感识别、匹配模型改进、深度特征融合等领域的最新进展。
1976

被折叠的 条评论
为什么被折叠?



