人工智能培训

搜索

论文代码开源:保持简单愚蠢(Keep it stupid simple)

[复制链接]
admin 发表于 2018-9-15 09:53:00 | 显示全部楼层 |阅读模式
admin 2018-9-15 09:53:00 627 0 显示全部楼层
人工智能论文代码开源:保持简单愚蠢(Keep it stupid simple)请注意该人工智能论文代码开源在github,大部分是python写的,框架可能是tensorflow或者pytorch。在本文中,我们提出了一种新颖的深度微词典学习和编码网络(DDLCN)。 DDLCN具有大多数标准深度学习层(汇集,完全,连接,输入/输出等),但主要区别在于基本卷积层被新的复合词典学习和编码层所取代。字典学习层学习输入训练数据的完整字典。在深度编码层,添加局部性约束以保证激活的字典基础彼此接近。接下来,激活的字典原子被组合在一起并传递到下一个复合字典学习和编码层。这样,第一层中的激活原子可以由第二字典中的深层原子表示。直观地说,第二个字典被设计为学习在inputdictionary原子之间共享的细粒度组件。以这种方式,可以获得字典原子的更具信息性和辨别力的低级表示。我们通过实际比较提出的DDLCN与几种字典学习方法和深度学习架构。四个流行的基准数据集的实验结果表明,与最先进的方法相比,拟议的DDLCN取得了有竞争力的结果。
In this paper, we propose a novel Deep Micro-Dictionary Learning and CodingNetwork (DDLCN).DDLCN has most of the standard deep learning layers (pooling,fully, connected, input/output, etc.) but the main difference is that thefundamental convolutional layers are replaced by novel compound dictionarylearning and coding layers.The dictionary learning layer learns anover-complete dictionary for the input training data.At the deep coding layer,a locality constraint is added to guarantee that the activated dictionary basesare close to each other.Next, the activated dictionary atoms are assembledtogether and passed to the next compound dictionary learning and coding layers.In this way, the activated atoms in the first layer can be represented by thedeeper atoms in the second dictionary.Intuitively, the second dictionary isdesigned to learn the fine-grained components which are shared among the inputdictionary atoms.In this way, a more informative and discriminative low-levelrepresentation of the dictionary atoms can be obtained.We empirically comparethe proposed DDLCN with several dictionary learning methods and deep learningarchitectures.The experimental results on four popular benchmark datasetsdemonstrate that the proposed DDLCN achieves competitive results compared withstate-of-the-art approaches.论文代码开源:保持简单愚蠢(Keep it stupid simple) zXQxPUxFfXXhUUFn.jpg
URL地址:https://arxiv.org/abs/1809.04185v1     ----pdf下载地址:https://arxiv.org/pdf/1809.04185v1    ----         ----github下载地址:https://github.com/Ha0Tang/DDLCN    ----    论文代码开源:保持简单愚蠢(Keep it stupid simple)请注意该人工智能论文代码开源在github,大部分是python写的,框架可能是tensorflow或者pytorch,keras,至于具体是哪一个没有完全测试。
回复

使用道具 举报

您需要登录后才可以回帖 登录 | 立即注册

本版积分规则 返回列表 发新帖

admin当前离线
管理员

查看:627 | 回复:0

快速回复 返回顶部 返回列表