文的自动分类 半监督学习 - 介于监督学习和无监督学习之间，算法: Graph Inference 或者Laplacian SVM 强化学习- 通过观察来学习做成如何的动作, 算法：Q-Learning以及时间差学习 机器学习－ 方法及流程 输入特征选择 – 基于什么进行预测 目标 – 预测什么 预测功能 – 回归、聚类、降维... Xn -> F(xn) -> T(x) 机器学习－ (NYU,2002), Facebook AI, Google Deepmind Theano (University of Montreal, ~2010), 学院派 Kersa, “Deep Learning library for Theano and TensorFlow” Caffe (Berkeley)，卷积神经网络,贾扬清 TensorFlow (Google) Spark MLLib

相对于机器学习，严重依赖于高端机，大量的 GPU 运算 • 高端机的成本高 开发工具 • 复杂的工具链 • 搭建环境花费时间 深度学习 à Azure Machine Learning 开发工具 à Azure Notebook Azure Machine Learning • 拥有不同运算性能的机器 • 降低成本，按需付费 • 支持不同的开源框架：TenserFlow、PyTorch、MXNet 等 Azure Azure Machine Learning SDK 2. 连接到 Azure Machine Learning 的 Workspace 3. 创建远程运算资源 4. 上传训练数据 5. 准备 training script 6. 把 training 任务提交到 Azure Machine Learning 1. 安装 Azure Machine Learning SDK 2. 连接到 连接到 Azure Machine Learning 的 Workspace 3. 创建远程运算资源 —— NC6 GPU machines 4. 上传训练数据 5. 准备 training script 6. 把 training 任务提交到 Azure Machine Learning 1. 创建 experiment 2. 创建 PyTorch estimator 3. 提交 training

Inference 机器学习服务 Amazon SageMaker Ground Truth Notebooks Algorithms + Marketplace Reinforcement Learning Training Optimization Deployment Hosting 人工智能服务 视觉 语音 语言 聊天机器人 预测 推荐 Personalize Forecast Lex Inference 机器学习服务 Amazon SageMaker Ground Truth Notebooks Algorithms + Marketplace Reinforcement Learning Training Optimization Deployment Hosting 人工智能服务 视觉 语音 语言 聊天机器人 预测 推荐 Personalize Forecast Lex Inference 机器学习服务 Amazon SageMaker Ground Truth Notebooks Algorithms + Marketplace Reinforcement Learning Training Optimization Deployment Hosting 人工智能服务 视觉 语音 语言 聊天机器人 预测 推荐 Personalize Forecast Lex

development are just some basic examples. Python is the language of choice for data analysis and machine learning, but it can also adapt to create games and work with embedded devices. Most importantly, it's all around the world. Many students learn Python as their first programming language. Many are learning it right now, many will learn it in the future. And for many of them, Python will be the only programming time. And that's typically how we run programs. We use the REPL for quick prototyping and for learning. On Linux and macOS a Python program can also be transformed into a shell script, by prepending

•Python的数据科学生态 •Python的数据科学生态 Kdnuggets&Kaggle :Deep Learning Framework Power Scores 2018 深度学习 在医疗领域应用实践 深度学习在医疗领域论文情况 Kwak G H J, Hui P. DeepHealth: Deep Learning for Health Informatics[J]. arXiv preprint arXiv:1909

Conv、Linear、Transformers 来源: Why GEMM is at the heart of deep learning, Pete Warden Convolution • AlexNet 模型推理各个层计算比例 • 86.1% • 2.6% 来源: Learning Semantic Image Representations at a Large Scale, Yangqing

Numpy and Scikit-learn With Jupyter Notebook 28 Python in Cloud AI Computing – Xilinx Machine Learning Suite Supported Frameworks: • Caffe • MxNet • Tensroflow Examples • DeepDetect REST Tutorial Intel x86 Supported FaaS • AWS F1 Xilinx Alveo PCIe 加速卡 • U50, U200, U250, U280 Ø 开发套件: Machine learning suite Ø AI inference IP: xDNN Ø AI development tools: xfDNN Ø 在云端的AI 应用，可以接受 python 语言进行部署，做数据预处理和后处理工作。

Numpy and Scikit-learn With Jupyter Notebook 28 Python in Cloud AI Computing – Xilinx Machine Learning Suite Supported Frameworks: • Caffe • MxNet • Tensroflow Examples • DeepDetect REST Tutorial Intel x86 Supported FaaS • AWS F1 Xilinx Alveo PCIe 加速卡 • U50, U200, U250, U280 Ø 开发套件: Machine learning suite Ø AI inference IP: xDNN Ø AI development tools: xfDNN Ø 在云端的AI 应用，可以接受 python 语言进行部署，做数据预处理和后处理工作。

Data + Machine Learning[R1] + Powerful Computation[R2] • 完全意义上的自下而上的方式 • 从海量的数据中去挖掘异构、动态、碎片化的知识 e.g., 从Web corpora、搜索日志等都可挖掘出有价值的知识 R1, http://www.erogol.com/brief-history-machine- learning/ R2, https://openai

Numpy and Scikit-learn With Jupyter Notebook 28 Python in Cloud AI Computing – Xilinx Machine Learning Suite Supported Frameworks: • Caffe • MxNet • Tensroflow Examples • DeepDetect REST Tutorial Intel x86 Supported FaaS • AWS F1 Xilinx Alveo PCIe 加速卡 • U50, U200, U250, U280 ➢ 开发套件: Machine learning suite ➢ AI inference IP: xDNN ➢ AI development tools: xfDNN ➢ 在云端的AI 应用，可以接受 python 语言进行部署，做数据预处理和后处理工作。