fa25-cs540 cs540
Lecture 27: Review Lecture 27: Review
Content locked
The catalog is public. Reading this content requires approved access.
Sign in to readLecture 27: Review
Final Exam Information
Natural Language Processing
Language Models
Training: The Markov Assumption
Unigram Model (k=0)
Bigram Model (k=1)
n-gram Model (k=n-1)
Issues and Solutions
Neural Networks
Perceptron
Multi-class Classification with Softmax
Gradient Descent
Minibatch Stochastic Gradient Descent (SGD)
Computational Graphs and Backpropagation
Numerical Stability
Gradient Vanishing with Sigmoid
Stabilizing Training
Generalization and Regularization
Training vs. Generalization Error
Underfitting and Overfitting
Convolutional Neural Networks (CNNs)
Motivation
2-D Convolution
Stride and Padding
Multiple Channels
Pooling Layers
CNN Architectures
Residual Connections
Recurrent Neural Networks (RNNs)
Architecture
Transformers
From RNNs to Transformers
Contextual Embeddings
Query, Key, and Value Matrices
Self-Attention Mechanism
Transformer Architecture
Uninformed Search
Breadth-First Search (BFS)
Depth-First Search (DFS)
Uniform-Cost Search
Iterative Deepening
Performance Comparison
Informed Search
Uninformed vs. Informed Search
A* Search
Games
Normal Form Games
Prisoner’s Dilemma
Dominant Strategies
Nash Equilibrium
Mixed Strategy Nash Equilibrium
Sequential-Move Games (Minimax)
Reinforcement Learning
Markov Decision Process (MDP)
Discounting
Bellman Equations
Q-Learning
Exploration vs. Exploitation
Q-Learning Algorithm
Practice Problems
Key Takeaways
Coming Up