白一帆的博客 | Ethan Blog

LeetCode 11. Container With Most Water

Array, Two Pointers

11. Container With Most Water Given n non-negative integers a1, a2, …, an , where each represents a point at coordinate (i, ai). n vertical lines are drawn such that the two endpoints of line i is...

Posted by baiyf on September 30, 2018

Python的异常处理与调试方法

exception handling and debugging method

Python的异常处理与调试方法异常即非正常状态，Python使用异常对象来表示异常，如果程序在编译或运行过程中发生错误，程序的执行过程就会发生变化，抛出异常对象。程序进入异常处理，如果异常对象没有被处理或者捕捉，程序就会执行回溯（Traceback）来终止程序。 Python标准异常异常名称描述 ...

Posted by baiyf on June 27, 2018

TensorFlow中常用的学习率递减方法

TensorFlow learning rate decay

Tensorflow中常用的学习率递减方法在训练神经网络时,控制学习率对训练的速度和准确度都有很大作用.逐渐减小学习率在实践中被证明对训练的收敛有正向效果,Tensorflow自带几种衰减方法: tf.train.piecewise_constant　分段常数衰减 tf.train.inverse_time_decay　反时限衰减 tf.train.polynomial_d...

Posted by baiyf on March 24, 2018

Sequence Models 第三周笔记

Sequence models & Attention mechanism

Sequence models & Attention mechanism Various sequence to sequence architectures Seq2seq model 机器翻译模型常常是encoder-decoder结构，也可以称为条件语言模型通常是经过编码器（绿色部分）把一句话转化为一个向量，再经过解码器输入这个向量并用beam searc...

Posted by baiyf on March 13, 2018

Sequence Models 第二周笔记

Natural Language Processing&Word Embeddings

Natural Language Processing&Word Embeddings Introduction to Word Embeddings Properities of word embeddings word embeddings可以解决类似man to woman is like king to __的问题,因为词义相近的词在向量上也比较接近 Cos...

Posted by baiyf on February 11, 2018

Sequence Models 第一周笔记

Recurrent Neural Network

Recurrent Neural Network 为了处理一句话的序列，首先面临的问题是怎么表示词word？ Representing word 表示word的一种方式是用字典的方式把每一个词word映射到一个唯一的向量vector，vector可以采用one-hot编码或者其他编码方式 Recurrent Neural Network 有一些循环神经网络RNN模型就可以完成上...

Posted by baiyf on February 11, 2018

Convolutional Neural Networks 第四周笔记

Face recognition&Neural style transfer

Face recognition&Neural style transfer Face recognition Verification&Recognition Verification：验证1对1的关系是否正确 Recognition：从数据库中识别出一个目标，是1对K的关系 One-shot学习 Learn from one example t...

Posted by baiyf on December 19, 2017

Convolutional Neural Networks 第三周笔记

Object detection

Object detection Object Localization 对于一个识别类别的神经网络 pedestrian car motorcycle background 同时输出￥$b_x,b_y,b_h,b_w$￥和类别标签(1-4) 定义标签$y = [P_c,b_x,b_y,b_h,b_w,C_1,C_2,C_3]$ $b_x,b_y$代表...

Posted by baiyf on November 28, 2017

Convolutional Neural Networks 第二周笔记

Case studies

Case studies Classical Networks LeNet - 5 input(32,32,1)—-6filters(5x5,s=1)—->conv1(28,28,6)—-avg pool(f=2,s=2)—->pool1(14,14,6)—-16filters(5x5,s=1)—->conv2(10,10,16)—-avg pool(f=2,s=2)...

Posted by baiyf on November 24, 2017

Convolutional Neural Networks 第一周笔记

Convolutional Neural Networks

Convolutional Neural Networks Computer Vision Problems:Image Classification;Object detection;Neural Style Transfer Edge Detection Vertical edge detection 垂直边缘filter \(\left[ \begin{matrix} ...

Posted by baiyf on November 22, 2017

Ethan Blog