Coursera Deep Learning Specialization(by Andrew Ng) 강의 정리
Week2: Neural Networks Basics
- Logistic Regression(로지스틱 회귀)
- Logistic Regression Cost Function(로지스틱 회귀 비용함수)
- Gradient Descent(경사하강법)
- Computation Graph & 미분
- Logistic Regression에서의 Gradient Descent
Week3: Shallow Neural Networks
- Neural network Representation and Vectorization
- Activation Function(활성화 함수)
- non-linear Activation Function를 사용하는 이유
- Activation Function(활성화 함수) 미분
- Weight Random Initialization(가중치 초기화)
Week4: Deep Neural Networks
- Forward Propagation
- Getting your Matrix dimension right
- Why deep Representations? (깊은 신경망이 더 많은 특징을 잡아내는 이유, 직관적으로)
- Parameters vs HyperParameters
- Forward and Backward propagation
Week1: Practical aspects of Deep Learning
- Train/Dev/Test sets
- Bias/Variance
- Basic Recipe for Machine Learning
- Regularization(L1,L2)
- Why Regularization reduces overfitting?
- Dropout Regularization
- Understanding Dropout
- Other regularization methods
- Normalizing Inputs
- Vanishing/Exploding Gradients
- Weight Initialize for Deep Networks
- Numerical Approximations of Gradients
- Gradient Checking
Week2: Optimization algorithms
- Mini Batch Gradient Descent
- Understanding Mini Batch Gradient Descent
- Exponentially Weight Average
- Bias Correction of Exponentially Weight Average
- Gradient Descent with Momentum
- RMSprop
- Adam Optimization Algorithm
- Learning Rate Decay
- The problem of Local Optima
Week3: Hyperparameter tuning, Batch Normalization
- Tuning Process
- Using an Appropriate Scale
- Normalizing Activations in a Network
- Fitting Batch Norm into Neural Networks
- Why Does Batch Norm Work?
- Batch Norm at Test Time
- Softmax Regression
- Training Softmax Classifier
Week1: ML Strategy(1)
- Orthonalization
- Single Number Evaluation Metric
- Satisficing and Optimizing Metrics
- Train/Dev/Test Set Distribusions
- Size of Dev/Test Sets
- When to change Dev/Test sets and Metrics?
- Why Human-Level Performance?
- Avoidable Bias(회피 가능 편향)
- Surpassing Human-Level Performance
- Improving Model Performance
Week2: ML Strategy(2)
- Carrying Out Error Analysis
- Cleaning Up Incorrectly Labelled Data
- Training and Testing on Different Distribusions
- Bias and Variance with Mismatched Data Distribusions
- Addressing Data Mismatch
- Transfer Learning
- Multi-task Learning
- What is end-to-end Deep Learning
- Whether to use end-to-end Deep Learning
Week1: Foundations of Convolutional Neural Networks
- Computer Vision
- Edge Detection Examples
- More Edge Detection
- Padding
- Stride
- Convolutions Over Volume
- One layer of a Convolutional Net
- Simple Convolutional Networks Example
- Pooling Layers
- Why Convolutions?
Week2: Deep Convolutional models(Case Studies)
- Why Look at Case Studies?
- Classic Networks (LeNet, AlexNet, VGG)
- ResNets
- Why ResNets Work?
- Networks in Networks and 1x1 Convolutions
- Inception Network Motivation
- Inception Network
- Transfer Learning
- Data Augmentation
- State of Computer Vision
Week3: Object Detection
- Object Localization
- Landmark Detection
- Object Detection
- Convolutional Implementation of Sliding Windows
- Bounding Box Predictions
- Intersection Over Union
- Non-max Suppression
- Anchor Boxes
- YOLO Algorithm
- Region Proposals