人工智能培训

搜索

论文代码开源:具有工作记忆模型的中国诗歌创作(Chinese Poetry Generation with a Working Memory Model)

[复制链接]
admin 发表于 2018-9-15 09:47:09 | 显示全部楼层 |阅读模式
admin 2018-9-15 09:47:09 1230 0 显示全部楼层
人工智能论文代码开源:具有工作记忆模型的中国诗歌创作(Chinese Poetry Generation with a Working Memory Model)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。我们提出了一个用于信号重建的原理贝叶斯框架,其中信号由基函数建模,其数量(和形式,如果需要)由数据本身确定。这种方法基于贝叶斯对传统稀疏重建和规范化技术的解释,其中通过贝叶斯模型选择通过先验施加稀疏性。我们演示了我们用于噪声1维和2维信号的方法,包括天文图像。此外,通过使用产品空间方法,可以将基函数的数量和类型视为整数参数,并且可以直接对其后验分布进行采样。可以看出,与单独计算贝叶斯证据相比,该技术可以实现计算效率的数量级增加,并且可以结合动态嵌套采样使用它进一步提高计算量。我们的方法可以很容易地应用于神经网络,其中它允许通过将节点和隐藏层的数量视为参数,由数据以原始贝叶斯方式确定网络架构。
We present a principled Bayesian framework for signal reconstruction, inwhich the signal is modelled by basis functions whose number (and form, ifrequired) is determined by the data themselves.This approach is based on aBayesian interpretation of conventional sparse reconstruction andregularisation techniques, in which sparsity is imposed through priors viaBayesian model selection.We demonstrate our method for noisy 1- and2-dimensional signals, including astronomical images.Furthermore, by using aproduct-space approach, the number and type of basis functions can be treatedas integer parameters and their posterior distributions sampled directly.Weshow that order-of-magnitude increases in computational efficiency are possiblefrom this technique compared to calculating the Bayesian evidences separately,and that further computational gains are possible using it in combination withdynamic nested sampling.Our approach can be readily applied to neuralnetworks, where it allows the network architecture to be determined by the datain a principled Bayesian manner by treating the number of nodes and hiddenlayers as parameters.论文代码开源:具有工作记忆模型的中国诗歌创作(Chinese Poetry Generation with a Working Memory Model) eOK7ooFlGlLlYUO7.jpg
URL地址:https://arxiv.org/abs/1809.04598v1     ----pdf下载地址:https://arxiv.org/pdf/1809.04598v1    ----         ----github下载地址:https://github.com/ejhigson/bsr    ----    论文代码开源:具有工作记忆模型的中国诗歌创作(Chinese Poetry Generation with a Working Memory Model)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:1230 | 回复:0

快速回复 返回顶部 返回列表