Machine Learning
homework代写 | report代写 | Machine Learning | 代做network – 这是利用pytorch进行训练的代写, 对Machine Learning的流程进行训练解析, 是比较典型的Machine Learning等代写方向, 这是值得参考的homework代写的题目
My sites/ 21S-STATS413-80/ Homeworks/ homework 01 --- NN & CNN Spring^2021 - Week 2
Spring 2021 - STATS413-80 - WU
Machine Learning
Homework 01 — NN & CNN
Detailed requirement
1. Read the following tutorial on pytorch: https://pytorch.org/tutorials/beginner/deep_learning_60min_blitz.html
2. Implement a classifier with Neural network by pytorch. Please follow: https://pytorch.org/tutorials/beginner/blitz/cifar10_tutorial.html#sphx-
glr-beginner-blitz-cifar10-tutorial-py. You need to try these settings in experiments:
Neural Network
Remove Conv layer in the tutorial.
Add more Linear layer. Compare result for 2-5 different linear layer.
Convolutional Neural Network
In tutorial, there are 2 conv layer. Change the number of "channel" in each layer (the first number in the input, the first input
channel is always 3 bacause an image have 3 channel: red, green and blue)
Decrease to 1 conv layer, increase to 3 conv layer. Compare results. Notice that when changing number of Conv, you need to
change the correspond kernel size and input channel for the later linear layer.
Keep the same number of layer, change the activation function. Try ReLU, LeakyReLU and Tanh. Compare the result.
Keep the tutorial network structure, change learning rate in optimizer definition. Try both smaller and bigger.
3. For the default setting on tutorial.
Print the training loss and testing loss in the same graph. x-axis is the epoch. Figure 1 is an example, make blue line to training loss and
green line to testing loss.
Print the training accuracy and testing accuracy in the same graph. x-axis is the epoch.
Visualize the classification result. There are 10 classes in the dataset. Print 20 example images in the testing dataset and label them true
class and predicted class. Highlight the wrong prediction. Hint: Use imshow(torchvision.utils.make_grid(images)) to print the example
images and make a separate table to label them. Calculate the accuracy for this 20 example and compare with the real testing accuracy.
4. For other compare method required. Write an table to report the final training loss, testing loss, training accuracy and testing accuracy.
5. Write the report. Follow the guidance in the template. (If there are any discrepancy between template and this detailed assignment. Please
follow this detail assignment.
Grading (total 100pts)
Homework report (90pts)
Neural Network (30pts)
introduction: 5pts
main equation: 5pts
algorithm: 5 pts
experiments 1: remove conv: 10pts
p p
experiments 2: compare 2-5 layer: 5pts
Convolution NN (60pts)
introduction: 5pts
main equation: 5pts
algorithm: 5 pts
Default setting: 25pts (completion: 10; loss-graph: 5; accuracy: 5; visualize: 5)
experiment 1-4: 5*4 = 20pts
Submit the code (py or ipynb): 10pts
Submission
Submit 2 files. DO NOT INCLUDE pdf in the zip. (1) One pdf report. It should use latex template provided on CCLE and follow guidence in the template. (2) One zip file includes every code you write including but not limited to ipynb file, py file, output plot as image if required. Please do not include dataset, code provided by TA.
example 1:
Submission status
Submission status No attempt
Grading status Not graded
Due date Sunday, 18 April 2021, 11:59 PM PDT
Time remaining 7 days 22 hours
Get the mobile app
Last modified -
Submission
comments Comments (0)
Add submission
You have not made a submission yet.
Homework guideline
Jump to...
2021 UC Regents Contact About Privacy Copyright UCLA links | UCLA Registrar MyUCLA Accessibility Couns/PsychSvc (CAPS)