90分钟,两个部分,完整的原题,求各位大佬加米!
(一)11题数理题选5道
1. You are waiting at a bus station for a bus; Bus arrives as a Poison process at a rate lambda; What is the distribution of wait time before the first bus arrives? What is the distribution of wait time before N's bus arrives?
2. What is the 1000th moment of a standard normal random variable?
3. X, Y, Z are independent and identically distributed conforming to a uniform distribution between (0,3); What is the probability that the median of X, Y, Z is between 1 and 2.
4. We have a random variable that conforms to a normal distribution N(mu, 1). Now we draw a data point × from this distributio
n. Using current information, how do we test the null hypothesis that mu = 5?
5. There are two red balls and 1 blue ball in a box; Randomly select a ball from the box and replace it with a blue ball. By the fi rst time when the box is full of blue balls, what is the expected number of replac calls. Pot 88.5bb. Flop 689ccd. You have 457.5bb left. What would you do here and why?
A. Check
B. Bet ⅓ pot (30bb)
C. Bet ⅓ pot (44.5bb)
D. Bet ⅘ pot (68bb)
E. All in
F. Some combinations of previous options (exactly what percentage would you mix them with?)
(二)10题简答题选5题道
1. 训练过程中,若一个模型不收敛,那么是否说明这个模型无效?导致模型不收敛的原因有哪些?
2. 什么是梯度消失?
3.请你简要讲一下,当训练集中类别不均衡时,哪个参数的结果最不准确?如何解决?
4.什么是卷积?
5. 梯度爆炸会带来什么问题?如何确定是否出现了梯度爆炸?
6.请简要介绍下tensorflow的计算图
7.解释一下神经网络中的Epoch、lteration、Batchsize(可以举例介绍)
8.数据预处理阶段,对数值特征归一化或标准化,理论上不会对哪种模型产生很大影响。
A 决策树; B k-means; C KNN
9. Adam 是什么?
10.dropout 在训练和测试过程中如何操作? |