摘要: ref: Broadcasting Tensorflow - Learning Tensorflow When we operate on arrays of different dimensionality, they can combine in different ways, either e 阅读全文
posted @ 2021-03-11 16:13 awr2sdgae 阅读(120) 评论(0) 推荐(0)
摘要: Visualizing A Neural Machine Translation Model Visualizing A Neural Machine Translation Model (Mechanics of Seq2seq Models With Attention) – Jay Alamm 阅读全文
posted @ 2021-03-09 15:00 awr2sdgae 阅读(57) 评论(0) 推荐(0)
摘要: def embedding(inputs, vocab_size, num_units, zero_pad=True, scale=True, scope="embedding", reuse=None): '''Embeds a given tensor. Args: inputs: A `Ten 阅读全文
posted @ 2021-03-09 14:01 awr2sdgae 阅读(88) 评论(0) 推荐(0)
摘要: An overconfident model is not calibrated and its predicted probabilities are consistently higher than the accuracy. For example, it may predict 0.9 fo 阅读全文
posted @ 2021-03-09 13:57 awr2sdgae 阅读(103) 评论(0) 推荐(0)
摘要: 由于生产环境目前还以1.13版本为主, 而google为了推进2.x已经下架了大部分官方教程, 这里简单做一些记录 1. name_scope / variable_scope 注意 tf.get_variable 不受 name_scope 影响 https://blog.csdn.net/u01 阅读全文
posted @ 2021-02-22 00:48 awr2sdgae 阅读(183) 评论(0) 推荐(0)
摘要: vscode 下默认的 python代码补全性能捉急, 使用 pylance替换默认的补全插件 https://marketplace.visualstudio.com/items?itemName=ms-python.vscode-pylance 阅读全文
posted @ 2021-02-22 00:45 awr2sdgae 阅读(130) 评论(0) 推荐(0)
摘要: 二分类模型做了3个实现 1. tensorflow lower API 实现逻辑回归二分类2. tensorflow senior API 实现二分类(sigmod函数由API内部默认实现)3. sklearn 的逻辑回归包用于比较输出 需要注意的是 tensorflow 中对于函数输出 nan 与 阅读全文
posted @ 2021-02-14 18:24 awr2sdgae 阅读(99) 评论(0) 推荐(0)
摘要: 参考: A visual proof that neural nets can compute any function https://www.cnblogs.com/yeluzi/p/7491619.html 阅读全文
posted @ 2021-01-15 16:58 awr2sdgae 阅读(213) 评论(0) 推荐(0)
摘要: 一个简单的方案是: 1. 通过大量正常文本, 统计 "字->字" 条件概率方阵, (例子中使用了26个英文字母+空格, 共27个基础字符, 构造的概率转移方阵为27*27) 2. 对于需要判断的文本, 将其拆解为顺序的2-gram词组, 对于每个词组, 查表得到第一个字符后出现第二个字符的概率, 3 阅读全文
posted @ 2021-01-11 21:01 awr2sdgae 阅读(1043) 评论(2) 推荐(0)
摘要: a-zero-math-introduction-to-markov-chain-monte-carlo-methods https://towardsdatascience.com/a-zero-math-introduction-to-markov-chain-monte-carlo-metho 阅读全文
posted @ 2020-12-28 22:07 awr2sdgae 阅读(352) 评论(0) 推荐(0)