Deep learning, as one of the three pillars of next-generation AI--the other two being big data and cloud computing, is a critical theoretical component for implementing a plethora of machine learning and AI technologies, ranging from Autopilot to ChatGPT.
Simultaneously, the innovation of Python has grown dramatically in the last decade. You can use Python in statistics, website development, and facial recognition. On top of that, Python is also used in groundbreaking areas of IT, like machine learning and data science. And it makes Python one of the most practical tools for deep learning implementation.
In this repository, I only used Python to demonstrate the fundamentals of deep learning, with no assistance from third-party frameworks like TensorFlow or PyTorch. It is your first step toward deep learning. Initially, I wrote these notebooks and inline annotations in Chinese for my studies. And I'll keep this page updated with relevant notes for each topic. Please keep checking back later.
You can find the datasets in the notebooks by this link.
- Python Base for Deep Learning
- Perceptron Neural Network
- Extension: Perceptron Explained Using Python Example
- Extension: Multi-layer Perceptron with Titanic Dataset
- Activation Functions
- Three-Layer Neural Networks
- Softmax Activation Function
- Batch Size
- Loss Functions
- Gradient Descent
- Learning Algorithms
- Backpropagation
- Backpropagation in Affine and Softmax Layers
- Backpropagation in Two-Layer Neural Networks
- Gradient Descent Stochastic and Momentum
- Gradient Descent Optimization
- Gradient Descent Optimization with MNIST
- Weights and Biases initialization
- Weights and Biases initialization with MNIST
- Batch Normalization
- Underfitting and Overfitting
- Weight Decay
- Dropout Algorithm
- Implementing Convolutional Neural Networks
References: