我想記錄我在 coursera 學習由 deeplearning.ai 提供的 Tensorflow 課程記錄 !

影片網址在 https://www.coursera.org/specializations/tensorflow-in-practice

由 Google 的 Laurence Moroney 講課，一共分成四大部分:

• Introduction to TensorFlow for Artificial Intelligence, Machine Learning, and Deep Learning
• Convolutional Neural Networks in TensorFlow
• Natural Language Processing in TensorFlow
• Sequences, Time Series and Prediction

Catalog

• Course 1: Introduction to TensorFlow for Artificial Intelligence, Machine Learning, and Deep Learning
  • Week 1: A New Programming Paradigm
  • Week 2: Introduction to Computer Vision
  • Week 3: Enhancing Vision with Convolutional Neural Networks
  • Week 4: Using Real-World Images
• Course 2: Convolutional Neural Networks in TensorFlow
  • Week 1: Exploring a Larger Dataset
  • Week 2: Augmentation: A technique to avoid overfitting
  • Week 3: Transfer Learning
  • Week 4: Multiclass Classifications
• Course 3: Natural Language Processing in TensorFlow
  • Week 1: Sentiment in text
  • Week 2: Word Embeddings
  • Week 3: Sequence models
  • Week 4: Sequence models and literature
• Course 4: Sequences, Time Series and Prediction
  • Week 1: Time series examples
  • Week 2: Deep Neural Networks for Time Series
  • Week 3: Recurrent Neural Networks for Time Series
  • Week 4: Real-world time series data

Introduction to TensorFlow for Artificial Intelligence, Machine Learning, and Deep Learning

第一周先從最基礎的 Tensorflow tools & syntax 學起

1. 從建立最簡單的神經網路開始來寫一個 regression
2. 使用神經網路來解 Computer Vision 問題! 實作 MNIST 和 FASHION MNIST 這兩個辨識數字和衣服的 dataset
3. 進階使用 CNN 的架構來建立神經網路模型
4. 最後用 CNN 來辨識真實世界的圖片

深入筆記在這邊: Course 1: Introduction to TensorFlow for Artificial Intelligence, Machine Learning, and Deep Learning

Convolutional Neural Networks in TensorFlow

第二周將更深入的了解 Tensorflow 中如何使用 CNN 模型來解決電腦視覺問題

1. 處理更加雜亂的真實世界圖片 (圖片有不同大小、顏色)
2. 解決模型 overfitting 的問題: 利用 augmentation 和 dropout
3. 使用 transfer learning 來獲得神力、加速學習
4. 學會將 CNN 應用於 Multiclass classification

深入筆記在這邊: Course 2: Convolutional Neural Networks in TensorFlow

Natural Language Processing in TensorFlow

第三周我們從原本的圖片，轉換跑道來處理文字，學習如何表達文字、並將文字來餵入模型、甚至創作文學 !

1. 學習怎麼讓要餵入模型的文字之間有意義
2. 進行文字的預處理 (tokenize, padding, embedding)
3. 掌握各種 sequence model (RNN, LSTM, GRU)
4. 使用 sequence model 與 text generation 來創造出莎士比亞的詩歌

深入筆記在這邊: Course 3: Natural Language Processing in TensorFlow

Sequences, Time Series and Prediction

第四周將介紹什麼是 time series，從非 ML 的統計方式到 DL 方式來處理 time series

1. Time series 介紹、用統計學模型來預測走向
2. 處理 time series 的資料以便丟進模型訓練，用最簡單的 DNN 來訓練並預測 time series
3. 用 RNN, LSTM 來處理 time series 問題，接觸 lambda layer
4. 加入 CNN 來處理 time series 問題，處理太陽黑子的真實資料集

深入筆記在這邊: Course 4: Sequences, Time Series and Prediction

Bonus

我還想學 Attention, Transformer, GANs, Auto-Encoder 等進階 DL 技巧 !!!

進階教材:

• aymericdamien / TensorFlow-Examples
• dragen1860 / TensorFlow-2.x-Tutorials
• ageron / handson-ml2