Views: 788 Author: Site Editor Publish Time: 2023-02-20 Origin: Site
| AI Engineer Camp | ||
| theme | Introduction | class |
| Introduction to Artificial Intelligence | ▫ What is artificial intelligence ▫ The history of artificial intelligence ▫ Common artificial intelligence products in life ▫ Common misunderstandings about artificial intelligence | 1 |
| The path to advancement as an algorithm engineer | ▫ Knowledge system of algorithm engineers ▫ Learning methods and materials ▫ How to read an AI paper | 1 |
| Machine Learning Basics (Ⅰ) | The technology behind artificial intelligence: AI models ▫ What is an AI model? ▫ Life cycle of AI model ▫ Judgment of AI model quality | 1 |
| Fundamentals of Machine Learning (Ⅱ) | Several ways to use AI models ▫ Using models ▫ Use training code ▫ Use training code + pre-trained model ▫ Use algorithms | 1 |
| Experience with Cognitive Services AI model | ▫ Get to know Microsoft Cognitive Services (ASR, TTS) ▫ Service overview + introduction to common tasks ▫ The service calls the method to obtain the result (eg audio file and viseme information) | 3 |
| Machine Learning Basics (Ⅲ) Understand model training and optimization algorithms | Understand model training and prediction through linear regression ▫ Model function of linear regression ▫ Training objectives of linear regression ▫ Prediction results of linear regression Understanding gradient descent through linear regression ▫ From model function to objective function ▫ Optimization objective function ▫ Gradient descent algorithm ▫ Implementation of gradient descent algorithm | 2 |
| Fundamentals of Machine Learning (IV) Model training environment | ▫ Stand-alone training ▫ Al Platform | 1 |
| Supervised model (Ⅰ) Two classification model | Understanding Logistic Regression: Regression Model for Binary Classification ▫ Model function ▫ Source of model function ▫ Advantages of two categories Objective functions and optimization objects of logistic regression ▫ Objective function ▫ Gradient descent process ▫ Implementation of logistic regression gradient descent ▫ BGD, SGD and MBGD From binary classification to multi-classification | 1 |
| Supervised model (Ⅱ) Other classification models | Introduce the basic principles, characteristics and selection methods of the model: ▫ Naive Bayes ▫ Decision tree ▫ SVM ▫ Boost tree (optional) | 2 |
| Write a gradient descent program for training Linear regression and logistic regression models | 2 | |
| Supervised model (Ⅲ) sequence prediction model | ▫ Markov chain, Markov process ▫ CRF & HMM principle ▫ Voice recognition ▫ Automatic music composition ▫ Automatic writing ▫ Entity recognition | 2 |
| How to use Sklearn | 1 | |
| Data Engineering (1) | ▫ Data cleaning ▫ Data normalization ▫ Feature extraction ▫ Encoding | 1 |
| Train a classification model | ▫ Use classification models for text classification | 2 |
| CRF actual combat | ▫ Train a NER model | 2 |
| Unsupervised models: clustering, etc. | ▫ Kmeans et al. ▫ Density clustering ▫GMM ▫PCA | 1 |
| Clustering algorithm in practice | ▫ Clustering | 2 |
| Fundamentals of Deep Learning (Ⅰ) | The history and current situation of neural networks fully connected neural network ▫ Hierarchy ▫ Elements ▫ Training objectives of neural network ▫ Inference of neural network ▫ Use gradient descent algorithm to train neural network Forward propagation and back propagation | 2 |
| from scratch Write a program and train the FC network | 3 | |
| Fundamentals of Deep Learning (Ⅱ) | Learn more about neural networks ▫ CNN ▫ RNN & LSTM ▫ Attention & Transformer Neural network applications ▫ Image ▫ Voice etc. | 1 |
| Fundamentals of Deep Learning (Ⅲ) | Model training method: ▫ Data preprocessing ▫ Weight initialization ▫ Optimization algorithm ▫ Regularization ▫ Learning rate and early stopping | 2 |
| artificial intelligence system | ▫ The current artificial intelligence system - deep learning system ▫ Resource requirements of deep learning systems (computing resources, data resources, human resources) ▫ History and design key points of deep learning framework | 2 |
| Deep learning training framework Tensorflow | What is a deep learning framework The role of deep learning framework ▫ Use of tensorflow 2.0 | 1 |
| Using Tensorflow Training multi-layer FC network | 2 | |
| NLP Basics (Ⅰ) | ▫ What is Natural Language Processing (NLP) ▫ Natural language processing tasks ▫ Model-based natural language processing | 1 |
| NLP Basics (Ⅱ) | ▫ word2vec ▫ bert | 2 |
| Text Classification Applications and Neural network solutions | ▫ Feature construction methods for text classification: bag-of-words model, TF-IDF, word2vec, etc. ▫ Text classification based on deep learning model: CNN + FC, BERT + FC, etc. ▫ Text classification data preparation: data annotation ▫ Text classification application: email classification (two classifications) vs. news classification (multiple classifications) | 2 |
| Data Engineering (2) How to convert text to vector | embedding: ▫ How embedded models evolve ▫VSM ▫ Topic Model ▫ word embedding ▫ contextual embedding ▫ large language model | 1 |
| Using word2vec and BERT Perform text embedding | ▫ Construct a word2vec model ▫ BERT pre-training ▫ Use external libraries to read BERT pre-trained models | 2 |
| text similarity problem solution | ▫ Application scenarios of text similarity tasks (intelligent dialogue) ▫ chart level ▫ word level ▫ sentence level ▫ paragraph level | 1 |
| Text similarity calculation Performance issues | ▫ Performance issues in machine learning ▫ Model distillation ▫ Quantification ▫ Pruning | 2 |
| Intelligent dialogue system and virtual Anchor platform | ▫ Intelligent dialogue system architecture ▫ Natural language understanding module ▫ Knowledge Base ▫ Dialogue management engine | 2 |
| Build a simple dialogue system | ▫ Use classification models and similarity models to identify user intentions respectively ▫ Return the corresponding answer according to the intention | 3 |
| Introduction to Knowledge Graph | ▫ What is knowledge graph ▫ Data model of knowledge graph ▫ Construction, storage and application of knowledge graph | 1 |
| knowledge extraction technology | ▫ Entity extraction ▫ Relation extraction ▫ Joint extraction | 2 |
| DialogEngine Instructions | 1 | |
| Virtual Anchor Instructions | 1 | |
| SmartKG How to use | 1 | |
| Knowledge graph cutting-edge technology | ▫ Research and application hotspots of knowledge graphs ▫ Knowledge graph application cases | 1 |
| reinforcement learning | ▫ Reinforcement Learning Concepts ▫ DQN ▫ Actor_Critic model | 2 |
| Reinforcement learning in action | ▫ Let the computer play Snake by itself | 2 |
| Generative Adversarial Network (GAN) | ▫ The concept of generative adversarial networks ▫ Evolution of the model ▫ Application | 2 |
| GAN actual combat | 2 | |
course features
Multi-field coverage: Covers multiple fields such as Python programming, data analysis, artificial intelligence, automated office, algorithms and data structures, etc., providing students with comprehensive and practical skills.
Cutting-edge technology learning: Focus on cutting-edge fields such as artificial intelligence, virtual character design, and AI generation technology to cultivate students’ innovation capabilities in the digital age.
Improvement of practical skills: Through actual project practice, case analysis, etc., the emphasis is on practical operation and problem solving, so that students can skillfully apply the knowledge they have learned.
Cultivation of problem-solving thinking: Focus on the teaching of algorithms and data structures, cultivate students' thinking methods of analyzing and solving problems, and improve programming abilities.
Comprehensive ability improvement: The knowledge system learned is interdisciplinary, so that students have the ability to comprehensively and comprehensively respond to complex technical and business challenges.
