These are all my assignments of the module CS5242 and there are one warm up and three assignments.
Warm up task is not graded, for another three assignment, and each one is 8% of full mark.
The structure of code: CS5242 Neural Network and Deep Learning --- Warm Up|---------data |---------util.py
--- Assignment1|-----code|----checkgradient.py
|----neuralnet_layers.py
|----neuralnet_model.py
|----question1.py
|----train_neuralnet.py
|-----data|----question_1
|----question_2_1
|----question_2_2_b
|----question_2_2_c
|-----Question_1
|-----Question_2
|-----reference
--- Assignment2|-----part1|------data
|------source|------problem1.py
|------warmup.py
|------assign2_utils.py
|------A0174365Y.pkl
|-----part2|------data
|------source|------assign2_utils_p2.py
|------cifar10.py
|------example_network.py
|------problem2_task1.py
--- Assignment3|-----filter_data
|-----part1|-------part1.py
|-----part2|-------part2.py
|-----part3|-------part3.py
|-----part4|-------part4.py
--- Project|-----data
|-----src|-------calculate_top_10_acc.py
|-------kdtree.py
|-------model_3DCNN.py
|-------model_lstm.py
|-------model_MLP.py
|-------MyFlatten.py
|-------preprocess_test.py
|-------preprocess_train.py
|-------read_pdb_file.py
|-------visualization.py
|-------vote.py
|-----figs
|-----project_description.pdf
There are some logs about my assignments as follows:
8.30 Complete the warm up task. After watching the video on coursera, try to build my first neural network without hidden layer.
The reference link : https://www.coursera.org/learn/neural-networks-deep-learning/notebook/NI888/planar-data-classification-with-a-hidden-layer
8.31 Complete assignment1 question1. use given w and b value of first neural network to calculate the w and b value of
the second neural network without hidden layers.
9.1 Build my first deep neural network with more than one hidden layers. Fix some bugs in neural network 1.
9.2 Complete all neural network (14-100-40-4,14-628-4 and 14-1428-4).
Use the weights and bias given to calculate the gradient of w and b, and evaluate the result with script.
Collect the performance data of three neural networks, including loss and accuracy. use matplotlib to plot and show results.
9.3 Modify the code, run code many times and collect experimental results.
Complete the report of assignment 1.
9.19 Complete problem1 of the assignment2.
9.20 Complete problem2 task1.
For mnist data set, got accuracy 93.29%. For cifar10 data set, got accuracy 43.11%. This is the baseline for model.
9.22 Complete problem2 task2.
Improved accuracy for mnist data is 98.2%. for cifar10 data set, improved accuracy is 55%.
9.23 Write the report for both problem.
10.11 Complete assignment 3 part 1
10.12 Complete assignment 3 part234
10.21 Complete assignment 3 report
11.2 Complete final project