代做EMATM0061: Statistical Computing and Empirical Methods, TB1, 2024 Assignment 9 代写数据结构程序

Assignment 9

EMATM0061: Statistical Computing and Empirical Methods, TB1, 2024

Introduction

This is the 9th assignment for Statistical Computing and Empirical Methods. This assignment is mainly based on Lectures 25, 26, 27 (see the Blackboards). Please  note that you don’t need to submit this assignment.

Load the tidyverse package:

library(tidyverse)

Some questions require the “glmnet” and “QSARdata” packages. If they are not already installed on your computer, please install them using “install.packages()” first.

1. Linear discriminant analysis

(Q1)

Describe the probabilistic model that underpins linear discriminant analysis.

(Q2)

In this question, we will train a linear discriminant analysis model to carry out the classification task to predict whether a hawk belongs to either the “Sharp-shinned” (SS) or the “Cooper’s” (CH) species of hawks, based on a four dimensional feature vector containing the weight, and the lengths of the wing, the tail and the hallux, which are generated by the following code.

library(Stat2Data) data(Hawks) hawks_total <- Hawks %>% select( Weight, Wing, Hallux, Tail, Species) %>% filter(Species=='SS' | Species =='CH') %>% drop_na() %>% mutate(Species=as.numeric(Species=='SS'))

Assume that we have the following train-test split of our dataset.

num_total <- hawks_total %>% nrow() # number of penguin data num_train <- floor(num_total*0.6) # number of train examples num_test <- num_total-num_train # number of test samples set.seed(0) # set random seed for reproducibility test_inds <- sample(seq(num_total),num_test) # random sample of test indicies train_inds <- setdiff(seq(num_total),test_inds) # training data indicies hawks_train <- hawks_total %>% filter(row_number() %in% train_inds) # train data hawks_test <- hawks_total %>% filter(row_number() %in% test_inds) # test data

Now, train a linear discriminant analysis model to carry out the classification task described above. Compute and report the train error and the test error.

(Q3) (*Optional) Implement your own linear discriminant analysis model. You should write a function called “generate_lda_model” which has two arguments, namely “training_data” (a data frame containing the training data) and “y_col_name” (a string specifying the name of the column that presents the labels). The remaining  columns of “training_data” represents the feature vectors. The function “generate_lda_model” returns a function that is your linear discriminant analysis model trained on the training data.

2. Logistic regression

(Q1) Describe the probabilistic model which underpins logistic regression.

(Q2)

Recall that the sigmoid function S: ℝ → (0,1) is defined by s(z) = 1/(1 + eZ). Generate the following plot which displays the sigmoid function:

(Q3) Now train a logistic regression model to predict whether a hawk belongs to either the “Sharp-shinned” or the “Cooper’s” species of hawks, based on a four-dimensional feature vector containing the weight, and the lengths of the wing, the tail and the hallux (similar to what you did in the Linear discriminant analysis question above. You can reuse the training test split in that question). Compute and report both the training error and the test error.

(Q4) (*optional)

Consider the following formula for the log-likelihood of the weights w ∈ ℝd and bias wo ∈ ℝ, given data D = ((x1, Y1), ⋯ , (xn, Yn)):

Show that and use it to demonstrate the following formulas for the derivatives:

Explain the role the above formula has in training a logistic regression model.

You can learn more about the glmnet approach to logistic regression here:

https://glmnet.stanford.edu/articles/glmnet.html#logistic-regression

3. Basic concepts in regularisation

Regularisation refers to the general technique within supervised learning of modifying an objective in some way so as to reduce the gap between test error and training error. Typically, this will increase the error on the training data. However, by reducing the gap between test and training errors, we can often improve performance (on unseen data).

Examples include

1)     l2 regularisation in the context of ridge regression for regression

2)    l1 or l2 regularised logistic regression for linear classification

(Q1)

Let’s review some key concepts relevant to regularisation. Write down your explanation of each of the following concepts.

1.      Hyper-parameter (and give an example of a hyper-parameter)

2.     Validation data

3.     The train-validation-test split

(Q2) What is the Euclidean (l2) norm and what is the l1 norm of a vector?

(Q3) The Ridge regression method and the Lasso method are both for learning a linear regression model from the data, by minimising an objective function. Describe what kinds of terms are included in their objective functions. What is the difference   between the two objective functions?

4. An investigation into ridge regression for high-dimensional regression

In this question we consider a high-dimensional regression problem. We consider a problem of predicting the melting point of a chemical compound from a relatively    high-dimensional feature vector of chemical descriptors.

To do this we shall use data from the ““QSARdata”” data library. Begin by checking if the ““QSARdata”” library has been installed.

Next load the ““QSARdata”” library and load the ““MeltingPoint”” data set.

library(QSARdata) data(MeltingPoint)

You will find a data frame called ““MP_Descriptors”” . The rows of the data frame correspond to different examples of chemical compounds and the columns correspond to various chemical descriptors. In addition you will find a vector called ““MP_Outcome”” which contains the corresponding melting point for each of the examples.

(Q1)

Begin by combining the data-frame of feature vectors ““MP_Descriptors”” together  with the column vector of melting points ““MP_Outcome”” . Combine these together into a single data frame entitled ““mp_data_total”” as follows.

mp_data_total<-MP_Descriptors %>% mutate(melting_pt=MP_Outcome)

How many variables are in your data frame? How many examples?

(Q2)

Next carry out a train-validate-test split of the ““mp_data_total”” data frame. You should use about 50% of the data to train the algorithm, about 25% to validate and about 25% to test

(Q3)

Our goal is to find a linear regression model φw,wo : ℝd → ℝ by φw,wo (X) = WXT +   Wo which estimates the melting point based upon the chemical descriptors. Create a function which takes as input training data (a matrix of features for training data and a vector of labels for training data), validation data (a matrix of features for validation data and a vector of labels for validation data) and a hyper-parameter λ . The function should train a ridge regression model using the specified value of λ, then compute the validation error and output a single number corresponding to the validation error. Your function should not be specific to this particular regression problem, but should apply to ridge regression problems in general. You can use the “glmnet” library within your function.

(Q4)

Next generate a sequence of candidate hyper-parameters called ““lambdas”” . The sequence should begin with 105 and increase geometrically in multiples of 1.25 and should be of length 70. That is, ““lambdas”” should contain the numbers

(Q5)

Now use your function to estimate the mean squared error on the validation data for a ridge regression model for the problem of predicting the melting point based on the chemical descriptors. Consider all of the hyperparameter values within your vector ““lambdas”” . Store the results of this procedure in a data frame.

(Q6)

Plot the validation error as a function of the hyper-parameter λ. Use the “scale x continuous()” function to plot the λ coordinate on a logarithmc scale.

Your plot may look like this:


(Q7)

Now use your results data frame. to determine the hyper-parameter λ with the lowest validation error. Retrain your ridge regression model with your selected value of the hyper-parameter λ and estimate the test error by computing the mean squared error on the test data.

(Q8) (*optional)

Does the test error computed above lead to a biased estimate of the mean squared error? Why can’t we use the mean squared error on validation data for the ridge regression model with the selected hyper-parameter as an estimate of the mean squared error on test data? Observe that the ridge regression model with the selected choice of λ has actually been trained twice: It was trained in order to compute the validation error, and then again to compute the test error. Comment on the computational efficiency of this procedure. Could it be easily improved? What memory implications would this have?


热门主题

课程名

mktg2509 csci 2600 38170 lng302 csse3010 phas3226 77938 arch1162 engn4536/engn6536 acx5903 comp151101 phl245 cse12 comp9312 stat3016/6016 phas0038 comp2140 6qqmb312 xjco3011 rest0005 ematm0051 5qqmn219 lubs5062m eee8155 cege0100 eap033 artd1109 mat246 etc3430 ecmm462 mis102 inft6800 ddes9903 comp6521 comp9517 comp3331/9331 comp4337 comp6008 comp9414 bu.231.790.81 man00150m csb352h math1041 eengm4100 isys1002 08 6057cem mktg3504 mthm036 mtrx1701 mth3241 eeee3086 cmp-7038b cmp-7000a ints4010 econ2151 infs5710 fins5516 fin3309 fins5510 gsoe9340 math2007 math2036 soee5010 mark3088 infs3605 elec9714 comp2271 ma214 comp2211 infs3604 600426 sit254 acct3091 bbt405 msin0116 com107/com113 mark5826 sit120 comp9021 eco2101 eeen40700 cs253 ece3114 ecmm447 chns3000 math377 itd102 comp9444 comp(2041|9044) econ0060 econ7230 mgt001371 ecs-323 cs6250 mgdi60012 mdia2012 comm221001 comm5000 ma1008 engl642 econ241 com333 math367 mis201 nbs-7041x meek16104 econ2003 comm1190 mbas902 comp-1027 dpst1091 comp7315 eppd1033 m06 ee3025 msci231 bb113/bbs1063 fc709 comp3425 comp9417 econ42915 cb9101 math1102e chme0017 fc307 mkt60104 5522usst litr1-uc6201.200 ee1102 cosc2803 math39512 omp9727 int2067/int5051 bsb151 mgt253 fc021 babs2202 mis2002s phya21 18-213 cege0012 mdia1002 math38032 mech5125 07 cisc102 mgx3110 cs240 11175 fin3020s eco3420 ictten622 comp9727 cpt111 de114102d mgm320h5s bafi1019 math21112 efim20036 mn-3503 fins5568 110.807 bcpm000028 info6030 bma0092 bcpm0054 math20212 ce335 cs365 cenv6141 ftec5580 math2010 ec3450 comm1170 ecmt1010 csci-ua.0480-003 econ12-200 ib3960 ectb60h3f cs247—assignment tk3163 ics3u ib3j80 comp20008 comp9334 eppd1063 acct2343 cct109 isys1055/3412 math350-real math2014 eec180 stat141b econ2101 msinm014/msing014/msing014b fit2004 comp643 bu1002 cm2030
联系我们
EMail: 99515681@qq.com
QQ: 99515681
留学生作业帮-留学生的知心伴侣!
工作时间:08:00-21:00
python代写
微信客服:codinghelp
站长地图