- 博客(9)
- 收藏
- 关注
原创 Dropout
不在最后一层使用dropoutIn lecture, we dicussed creating a variable d[1]d^{[1]}d[1] with the same shape as a[1]a^{[1]}a[1] using np.random.rand() to randomly get numbers between 0 and 1. Here, you will use a vectorized implementation, so create a random matrix $D
2020-10-18 11:41:33 288
原创 L2 Regularization
L2 RegularizationThe standard way to avoid overfitting is called L2 regularization. It consists of appropriately modifying your cost function, from:J=−1m∑i=1m(y(i)log(a[L](i))+(1−y(i))log(1−a[L](i)))(1)J = -\frac{1}{m} \sum\limits_{i = 1}^{m} \large{(}
2020-08-16 10:59:31 250
原创 He initialization
Initialization:一个好的初始化可以:Speed up the convergence of gradient descentIncrease the odds(胜算,几率) of gradient descent converging to a lower training (and generalization) error初始化的shape:数据因当是(样本内容,样本数量)第一个参数W的设置因当是(样本数量,本层神经元数)Wx的设置为(输入层数,本层)He initiali
2020-08-16 09:35:44 535
原创 Day 5(Improving Deep Neural Networks: Hyperparameter tuning, Regularization and Optimization Week 1)
Algorithm Design决定:layershidden unitslearning ratesactivation functions训练:train sets, 70%dev sets, 20%test sets, 10%Bias and variancehigh bias = under fitting 高偏差 欠拟合hihg variance = over fitting 高方差 过度拟合train set error:dev set error:dev >
2020-08-14 10:10:13 138
原创 python特定代码段
将新数据放入字典并记数strs = ["flower", "flow", "flight"]dic = {}for i in strs: if i in dic.keys(): dic[i] = 1 else: dic[i] += 1print(dic)# {'f': 3, 'fl': 3, 'flo': 2, 'flow': 2, 'flowe': 1, 'flower': 1, 'fli': 1, 'flig': 1, 'fligh': 1, '
2020-07-30 19:34:24 100
原创 Day 1 (二进制搜索树,排序,数据结构对比)
二进制搜索树搜索的时间复杂度是lg(N),总层数是lg(N)的向下取整,N是总节点数。对数公式a^x = NlogaN = x排序:counting sort and radix sortcounting sort会浪费大量空间和时间 在数字跨度比较大的时候,因为他要对所有数位开一个位置radix sort通过将每一位数的counting sort来确定每一个数的位置python使用的是timsortdef counting_sort(A, d): L = [] # K
2020-07-30 12:29:14 201
原创 Hopefully, it is a start.
浑浑噩噩了大半年,在这个寒假过半的时候,感觉自己已经到了真正的触底反弹的时候了。上一周,每天平均使用十个小时的手机(看剧),chrome一共使用了27小时,B站一共使用了24小时,一周七天我有整整两天的时间花费在了看剧上,天哪,还有40分钟的网飞。真的要像自己学雅思那段时间一样逼一逼自己了。能有一个大的努力方向吗看看crousera的课 看看花书 看看算法的课 leetcode也不要停 汲取不同的知识 生活比电视剧更值得时间规划为8-22点,4+10,14个小时,每两个小时一个小段,一共7段,每结
2020-07-27 16:11:47 185
原创 Thought about everything in every day.
Thought about everything in every day.20.May.2020Python todayenumerate()When you used it?for i, data in enumerate (trianloder, 0):## trainloder: data file loader in Pytorchsource: https://www.runoob.com/python/python-func-enumerate.html>>>
2020-05-20 15:13:20 111
空空如也
空空如也
TA创建的收藏夹 TA关注的收藏夹
TA关注的人