Comments:
- *: The student's homeworks marked with * are those with more information not provided and taught by the teacher.
- #: The student's homeworks marked with # are those with references.
HWR6 - Reading 6: Backpropagation learning algorithm
Goal:
- Understand the MLP learning algorithm: backpropagation.
- Reading: "Chapter 1 Using neural nets to recognize handwritten digits", Neural Networks and Deep Learning, by Michael Nielsen, 2015.
- Focus more on the 2nd section (Sigmoid neurons) and 5th section (Learning with gradient descent).
- Your report should explain more on : (1) Why use sigmoid for MLP, (2) What is MNIST data set, (3) How to design an MLP to recognize digits, (4) What is backpropagation algorithm, (5) what is gradient descent.
- Add references into your report, if you read more information than that provided by the teacher.
- You should write your report in your Weebly site.
- Deadline: 2016/05/2 24:00.
HWR5 - Reading 5: MLP and XOR
Goal:
- Learn how MLP (multiple-layer perceptron) can solve XOR, a simple nonlinear classification problem.
- Readings: the same with HWR4.
- Your report should explain more on : (1) What is MLP, (2) what is the difference between Perceptron and MLP, (3) why MLP can solve XOR.
- Write a reading report in English. The length of the report is “about” 500 words. More than 500 words are welcome. Figures and illustrations are welcome.
- Add references into your report, if you read more information than that provided by the teacher.
- You should write your report in your Weebly site.
- Deadline: 2016/04/18 24:00.
HWR4 - Reading 4: Perceptron learning algorithm
Goal:
- Understand the learning algorithm of perceptron: the delta rule.
- Readings:
- "Chapter 11 Multilayer Perceptrons," Introduction to Machine Learning, 2nd, E. Alpaydin, MIT Press, 2010. (45 pages)
- "Chapter 4 Artificial Neural Networks", Machine Learning, T. Mitchell, McGraw-Hill, 1997. (46 pages)
- Write a reading report in English. The length of the report is “about” 500 words. More than 500 words are welcome. Figures and illustrations are welcome. More references are also welcome.
- Your report should explain more on : (1) What is perceptron by representing it with the formula of linear classifiers, and (2) the learning algorithm of perceptron. You should write the formula of the delta rule of the algorithm, and also write a pseudo code of the algorithm. Some figures to complement the explanation of the algorithm will be also good.
- Add references into your report, if you read more information than that provided by the teacher.
- You should write your report in your Weebly site.
- Deadline: 2016/04/04 24:00.
HWR3 - Reading 3: How does machine learning work: decision tree as an example
Goal:
Links: Steffi, Jonathan*, Spencer, Karissa, Lolly, Cindy, Jacky, Kirito*, Tommy, Dioxin*, Daniel.
- Learn an important method in machine learning: decision tree.
- Readings:
- HOW MACHINE LEARNING WORKS [INTERACTIVE], 2016/02/23. (中文)
- Undergraduate machine learning 31: Decision Trees, 2012. UBC, Prof. Nando de Freitas. Youtube:00:39:42. (SlideShare)
- Decision Tree, by Victor Lavrenko, Youtube 8 部影片, 2014/06/22.
- "Chapter 3 Decision Tree Learning", Machine Learning, T. Mitchell, McGraw-Hill, 1997. (28 pages)
- Write a reading report in English. The length of the report is “about” 500 words. More than 500 words are welcome. Figures and illustrations are welcome. More references in your report are also welcome.
- Your report should explain more on : decision tree, decision forest, and the relationship between decision tree and random forest. It is better if you can give some examples and explanations of learning algorithms of decision trees/forests, such as ID3, random forest, ...
- Add references into your report, if you read more information than that provided by the teacher.
- You should write your report in your Weebly site.
- Deadline: 2016/03/28 24:00.
Links: Steffi, Jonathan*, Spencer, Karissa, Lolly, Cindy, Jacky, Kirito*, Tommy, Dioxin*, Daniel.
HWR2 - Reading 2: Linear classifier and neural networks
Goal:
- Understand the relationship between linear classifier and a classical neural network: perceptron.
- Chapter 8 Classical Models of Neural Network, Celebi Tutorial: Neural Networks and Pattern Recognition Using MATLAB.
- Write a reading report in English. The length of the report is “at most” 500 words. Figures and illustrations are welcome. More references in your report are also welcome.
- Organize your web site by the suggestion of Tutorial 2.
- Your report should try to explain more on : basic definition of linear classifier, geometric interpretation of linear classifier, perceptron, relation of linear classifier and perceptron.
- You should write your report in your Weebly site. Organize your web site by the suggestion of Tutorial 2.
- Deadline: 2016/03/21 24:00.
HWR1 - Reading 1 : What is pattern recognition
Goal:
- Learn what is pattern recognition (PR) by reading the following online article.
- Chapter 1 Pattern Classification, Celebi Tutorial: Neural Networks and Pattern Recognition Using MATLAB.
- Write a reading report in English.
- Write a report with “at most” 500 words in English.
- Your report should try to explain more on : goal of PR, applications of PR, steps or algorithms of PR, fundamental concepts of PR.
- You should write your report in your Weebly site.
- Deadline: 2016/03/14 24:00.