GIT

Basics of git

• a stream of snapshots

  Git thinks of its data more like a series of snapshots of a miniature filesystem.

• nearly every operations is local

  We have the entire history of the project right there on our local disk, most operations seem almost instantaneous

• three states

  • committed means that the data is safely stored in your local database.
  • modified means that you have changed the file but have not committed it to your database yet
  • staged means that you have marked a modified file in its current version to go into your next commit snapshot

• three main sections of a Git project

  • working directory
  • staging area
  • .git directory(repository)

MULTI-SCALE CONTEXT AGGREGATION BY DILATED CONVOLUTIONS

空洞卷积

1. Introduction

Dense prediction

The goal is to compute a discrete or continuous label for each pixel in the image. A prominent example is to semantic segmentation , which calls for classifying each pixel into one of a given set of categories.

Semantic segmentation is challenging because it requires combining pixel-level accuracy with multi-scale contextual reasoning.

像素精度与上下文推理相结合

Modern image classification networks integrate multi-scale contextual information via successive pooling and subsampling layers that reduce resolution until a global prediction is obtained.

Dense prediction calls for multi-scale contextual reasoning in combination with full-resolution output.

A network module that aggregates multi-scale contextual information without losing resolution or analyzing rescaled image.The module can be plugged into existing architectures at any resolution.

is designed specifically for dense prediction

It is a rectangular prism of convolution layers, with no pooling or subsampling.

The module is based on dilated convolutions, which support exponential expansion of the receptive field without loss of resolution or coverage.

Deep Residual Learning for Image Recognition

Q：Is learning better networks as simple as stacking more layers？

A：Noooo

​ An obstacle to answering this question was the notorious problem of vanishing/exploding gradients , which hamper convergence from the beginning .

This problem has been largely addressed by normalized initialization and intermediate normalized layers , which enable networks with tens of layers to start converging for SGD with bp.

常规的网络的堆叠（plain network）在网络很深的时候，效果却却越来越差，train set和test set都变差。

产生这一问题的原因之一即 deeper network，gradients vanish越明显，network的训练效果也不是很好。

Resnet network的目标是在网络加深的情况下解决gradients vanish的问题。

U-Net：Convolutional Networks for Biomedical Image Segmentation

Abstract ：

The architecture consists of a contracting path to capture context and a symmetric expanding path that enables precise localization

ROC curve , AUC & Recall

ROC（Receiver Operating Characteristic）curve ：

In signal detection theory, a receiver operating characteristic (ROC), or simply ROC curve, is a graphical plot which illustrates the performance of a binary classifier system as its discrimination threshold is varied.

ROC curve 和 AUC curve常用来被评价一个二值分类器（binary classifier） ，输出结果只有两种类别的模型，当讯号侦测（或变数测量）的结果是一个连续值，类与类的边界必须用一个阈值（threshold） 来界定

横轴：伪阳率（FPR）False Positive Rate，在所有实际为阳性的样本中，被正确地判断为阳性的比率

纵轴：真阳率（TPR）True Positive Rate，在所有实际为阴性的样本中，被错误地判断为阳性的比率

A systematic study of the class imbalance problem in convolutional neural networks

• several methods：oversampling，undersampling，two-phase training，thresholding
• problem：some classes have a significantly higher number of examples in the training set than other classes –class imbalance ，Eg：computer vision，medical diagnosis，fraud detection，the frequency of one class can be 1000 times less than another class
• A new method for CNNs was introduced that trains the network in two-phases in which the network is trained on the balanced data first and then the output layers are fine-tuned

Question：

Determine whether an integer is a palindrome . Do this without extra space.

Example：

12321

Some hints：

Could negative integers be palindromes? (ie, -1)

If you are thinking of converting the integer to string, note the restriction of using extra space.

You could also try reversing an integer. However, if you have solved the problem “Reverse Integer”, you know that the reversed integer might overflow. How would you handle such case?

There is a more generic way of solving this problem.

Question：

Reverse digits of an integer.

Example1: x = 123, return 321

Example2: x = -123, return -321

Example3: x = 120, return 21

Note：Assume we are dealing with an environment which could only hold integers within the 32-bit signed integer range. For the purpose of this problem, assume that your function returns 0 when the reversed integer overflows.

Question：

Given an array of integers, return indices of the two numbers such that they add up to a specific target.

You may assume that each input would have exactly one solution, and you may not use the same element twice.

Example：

1
2
3
4

Given nums = [2, 7, 11, 15], target = 9,
Because nums[0] + nums[1] = 2 + 7 = 9,
return [0, 1].

支持向量机分类器（Support Vector Classifier）

根据训练样本的分布，搜索所有可能的线性分类器中最佳的那一个 ，决定其直线位置的样本并不是所有训练数据，而是其中的两个空间间隔最小的两个不同类的数据点 ，把这种可以用来真正帮助决策最优线性分类模型的数据点叫做“支持向量”。