lecture 3,Loss Function and Optimization
来源:互联网 发布:高分一号数据在哪下载 编辑:程序博客网 时间:2024/06/06 00:46
1,处理指数计算问题时,为了防止overflow 的处理方法(我还是存在疑问,这相当与求超级大的数的导数):
f = np.array([123, 456, 789]) # example with 3 classes and each having large scoresp = np.exp(f) / np.sum(np.exp(f)) # Bad: Numeric problem, potential blowup# instead: first shift the values of f so that the highest number is 0:f -= np.max(f) # f becomes [-666, -333, 0] p = np.exp(f) / np.sum(np.exp(f)) # safe to do, gives the correct answer
2,关于LR为凸函数,可以保证最值(中科大面试时被问到)???
3,svm的代码复现,看了很多遍svm的思路,也被面试问到过svm,不如尝试复现一次或者trvarse吧。。。
阅读全文
0 0
- lecture 3,Loss Function and Optimization
- CS231n系列课程Lecture 3: Loss functions and Optimization
- 【CS231n winter2016 Lecture 3 (Linear classification II/loss function/optimization/SGD)】
- CS231n-3-Loss Function & Optimization
- cs231n第三节Loss Function and Optimization
- stanford_CS231n_learning note_Lec_03 Loss function and optimization
- [cs231n之第三课]Loss function and optimization
- CS231n 2017Spring Lecture3 Loss Function and Optimization总结
- CS231N-Lecture3 Loss Functions and Optimization
- Loss Function
- Loss function
- loss function
- Differences between L1 and L2 as Loss Function and Regularization
- Lecture 3 Divide and Conquer
- Loss Function , Cost Function and Kernel Function in ML(To be continued)
- 损失函数(loss function)
- 损失函数(loss function)
- Huber Loss function
- c++知识点 --- 输出宽度及填充的使用
- 结对项目-数独程序扩展
- 给家庭理财者的十条忠告
- 【Photoshop】合并一系列序列帧图片成序列帧大图
- gdb调试器与make工程管理器
- lecture 3,Loss Function and Optimization
- C++ 数据类型
- 投资理财的十大经典电影
- Map的实现类
- svm代码复现
- Dom对节点操作
- 无序数组最小的K个数
- LintCode54: String to Integer II
- 乐趣:p++和**k