Course Objectives
1. Can explain the training of prediction machines and classifiers as well as backpropagation of errors.
2. Can implement a neural network with input, hidden, and output layers, capable of identifying handwritten digits.
Rubric
| Ideal Level | Standard Level | Unacceptable Level |
Principle | Can derive the matrix form for output layer computation and weight updates using the backpropagation method. | Understands and can explain the matrix form for output layer computation and weight updates using the backpropagation method. | Can provide a general explanation of the method for updating weights using the output layer computation method and the backpropagation method. |
Implementation in Python | Can identify one's own handwritten digits. | Can identify handwritten digits using the MNIST dataset. | Can implement a simple neural network. |
Assigned Department Objectives
学習・教育到達度目標 B-4
See
Hide
学習・教育到達度目標 D-1
See
Hide
Teaching Method
Outline:
Can explain classifiers and their learning methods based on neural networks, which are fundamental to deep learning (deep neural networks), and can implement them in Python.
Style:
Students are assumed to have already acquired knowledge of matrix inner products and basics of Python (functions, matrix operations using Numpy) through other classes or self-study.
【Lecture hours: 30 hours + Self-study hours: 60 hours】
Notice:
This course is designed for individuals who have acquired linear algebra and programming skills, aiming to learn the principles used in machine learning (artificial intelligence). There is a significant portion where matrix operations are carried out by hand, demanding not only proficiency in matrix operations but also "grit."
Characteristics of Class / Division in Learning
Course Plan
|
|
|
Theme |
Goals |
1st Semester |
1st Quarter |
1st |
Simple Prediction Machine |
Can explain methods for learning predictive machines.
|
2nd |
Learning a Simple Classifier |
Can explain how to use classifiers to categorize data.
|
3rd |
Neuron |
Can describe the structure of a neuron and networks formed by neurons.
|
4th |
Learning Weights from Two or More Nodes |
Can explain how to adjust the internal parameters of a neuron.
|
5th |
Backpropagation from Many Nodes |
Can explain how to adjust the internal parameters of a neural network with many nodes.
|
6th |
Backpropagation to Many Layers |
Can explain how to propagate errors from the output layer to the hidden layers.
|
7th |
Midterm Exam |
|
8th |
Weight Update |
Can explain the equations used for updating the weights of an entire neural network.
|
2nd Quarter |
9th |
Introduction to Python |
Can create Python programs using classes, among other things, to implement neural networks.
|
10th |
Definition of Neural Network Class 1 |
Can implement the structure of a neural network.
|
11th |
Definition of Neural Network Class 2 |
Can implement a neural network that is capable of backpropagation.
|
12th |
Learning the Network |
Can train a neural network and visualize the process.
|
13th |
Handwritten Digit Dataset 1 |
Can explain what the MNIST dataset of handwritten digits is.
|
14th |
Handwritten Digit Dataset 2 |
Can implement a neural network that recognizes handwritten digits.
|
15th |
|
|
16th |
|
|
Evaluation Method and Weight (%)
| Examination | Presentation | Mutual Evaluations between students | Behavior | Portfolio | Other | Total |
Subtotal | 100 | 0 | 0 | 0 | 0 | 0 | 100 |
Basic Proficiency | 0 | 0 | 0 | 0 | 0 | 0 | 0 |
Specialized Proficiency | 100 | 0 | 0 | 0 | 0 | 0 | 100 |
Cross Area Proficiency | 0 | 0 | 0 | 0 | 0 | 0 | 0 |