课程项目
学生将在 KTBlocks 系统上构建、测试和发布自己的游戏项目
编码平台
KTCoder 一体化编码平台支持我们的互动在线课程、专业化课程体系，以及学生对学习的热情。
学生答疑辅导时间
答疑辅导时间由我们高素质的助教团队主导。这是帮你的代码获取即时反馈的免费便捷途径。
进度报告
KTBYTE 将通过电子邮件的方式向家长发送学生的课堂表现和成绩报告
结业证书
学生完成每门课程后均可申请结业证书。
課程描述:
学习最新的监督式学习技巧， 该方法应用于面部识别、语音识别和自动驾驶汽车等常见应用 。本课程还将为学生提供具有GPU加速功能的Linux服务器来运行算法。主题包括回归，测试分类，卷积图像识别等等。
特点：
- 本课程为学生提供专业研究级硬件，包括计算服务器， 如具有128GB随机访问存储器的16核服务器， 数TB的快速存储和研究级图形处理器, 如Titan X Pascal GPU.
- 本课程还会教授学生如何使用linux工具和CUDA + GPU 加速Python研究环境， 使用Tensorflow-GPU 1.4, Keras 2.1.4等使用lubcudnn 6和7编译的工具。
- 课程材料来自最近2-5年发表的最新学术研究， 包括深度网络，单发检测，卷积或矢量化语言模型， 以及（如果时间允许）一些项目展示， 比如AlphaZero Go和GAN引发的序列到序列学习.
本课程不再使用Theano，学生将主要使用Tensorflow支持的Keras深度学习库进行建模。
Research Projects from KTBYTE students
入班要求
修完[CORE 5b]或AP计算机科学，或获得导师同意。还需掌握代数II数学。建议修[AI 1]但不强制。
Class Description:
学习最新的监督式学习技巧， 该方法应用于面部识别、语音识别和自动驾驶汽车等常见应用 。本课程还将为学生提供具有GPU加速功能的Linux服务器来运行算法。主题包括回归，测试分类，卷积图像识别等等。
特点：
- 本课程为学生提供专业研究级硬件，包括计算服务器， 如具有128GB随机访问存储器的16核服务器， 数TB的快速存储和研究级图形处理器, 如Titan X Pascal GPU.
- 本课程还会教授学生如何使用linux工具和CUDA + GPU 加速Python研究环境， 使用Tensorflow-GPU 1.4, Keras 2.1.4等使用lubcudnn 6和7编译的工具。
- 课程材料来自最近2-5年发表的最新学术研究， 包括深度网络，单发检测，卷积或矢量化语言模型， 以及（如果时间允许）一些项目展示， 比如AlphaZero Go和GAN引发的序列到序列学习.
本课程不再使用Theano，学生将主要使用Tensorflow支持的Keras深度学习库进行建模。
Prerequisites:
修完[CORE 5b]或AP计算机科学，或获得导师同意。还需掌握代数II数学。建议修[AI 1]但不强制。
Related Classes
Syllabus:
Machine Learning Review
Review concepts taught in our AI 1 course, including linear and logistic regression and basic models for classification and regression problems.
Introduction to Neural Networks
In this class we'll learn about the Perceptron as the building block for neural networks and deep learning
Introduction to TensorFlow
In this class we'll start exploring how to use the TensorFlow library.
More TensorFlow, Intro to Keras
In this class we'll keep working with TensorFlow and start learning to use Keras.
Working with Images
In this lesson we'll start learning how to process images with our ML architectures, including running a model for image ID or generation.
Convolutional Neural Networks (CNN)
Today we'll start exploring a new type of neural network and learn about regularizations
Transfer Learning
Transfer learning allows us to copy effective parts of existing models. Today we will also introduce the midterm project.
Midterm Project: Image Classification
Project Day
Finish Midterm Project, presentations
Students will present their work from the midterm project. Class discussion on topics to cover in Unit 2 in order to meet student goals.
Recurrent Neural Networks (RNNs) - Part 1
RNNs can be used to make predictions about time series data, like words in a sentence. This lesson focuses on the basics of building these models and how to avoid common issues.
Recurrent Neural Networks (RNNs) - Part 2
In this lesson we'll wrap up our discussion of RNNs and gated layers (GRU and LSTM). Then we'll cover a longer demo of text generation with RNNs.
Image Generation
In this lesson we'll delve into the theory of how deep learning can be used to generate images, and cover the basics of models like GANs and stable diffusion.
Research Project Brainstorming
In today's class we'll review topics covered in the course. Students will be provided with time to start brainstorming a research question for their final project.
Research Projects
Continue working on research projects.
Research Project Presentations
Come to class prepared to present your research project.
课程大纲
Machine Learning Review
Review concepts taught in our AI 1 course, including linear and logistic regression and basic models for classification and regression problems.
Introduction to Neural Networks
In this class we'll learn about the Perceptron as the building block for neural networks and deep learning
Introduction to TensorFlow
In this class we'll start exploring how to use the TensorFlow library.
More TensorFlow, Intro to Keras
In this class we'll keep working with TensorFlow and start learning to use Keras.
Working with Images
In this lesson we'll start learning how to process images with our ML architectures, including running a model for image ID or generation.
Convolutional Neural Networks (CNN)
Today we'll start exploring a new type of neural network and learn about regularizations
Transfer Learning
Transfer learning allows us to copy effective parts of existing models. Today we will also introduce the midterm project.
Midterm Project: Image Classification
Project Day
Finish Midterm Project, presentations
Students will present their work from the midterm project. Class discussion on topics to cover in Unit 2 in order to meet student goals.
Recurrent Neural Networks (RNNs) - Part 1
RNNs can be used to make predictions about time series data, like words in a sentence. This lesson focuses on the basics of building these models and how to avoid common issues.
Recurrent Neural Networks (RNNs) - Part 2
In this lesson we'll wrap up our discussion of RNNs and gated layers (GRU and LSTM). Then we'll cover a longer demo of text generation with RNNs.
Image Generation
In this lesson we'll delve into the theory of how deep learning can be used to generate images, and cover the basics of models like GANs and stable diffusion.
Research Project Brainstorming
In today's class we'll review topics covered in the course. Students will be provided with time to start brainstorming a research question for their final project.
Research Projects
Continue working on research projects.
Research Project Presentations
Come to class prepared to present your research project.