9. Topics for Exam 2
Topics for Exam 2
Foundations:
  • Python, NumPy, and Pytorch
  • Vector, matrix, and tensor operations
  • Equations of lines and planes in multi-dimensional space (hyperplanes)
Neural Network Basics:
  • Neuron model and network architectures
  • Single neuron and activation functions
  • Layers of neurons, weight matrices, and biases
  • Multi-layer neural network architecture
  • Decision boundaries and their relationship to hyperplanes
Computational Graphs:
  • Construction and interpretation of computational graphs
  • Forward pass and backpropagation (backward pass)
Loss functions: 
  • Mean Squared Error (MSE)
  • Mean Absolute Error (MAE)
  • Hinge loss (SVM)
  • Cross-entropy loss
PyTorch Workflow:
  • Data transforms
  • Datasets
  • Dataloaders
  • Model definition
  • Loss functions
  • Optimizers and learning rate schedulers
  • Training loop
  • Evaluation loop
  • Performance metrics and visualization (plots)
  • Saving/loading a checkpoint
Convolutional Neural Networks (CNN):
  • Creating and analyzing CNNs using PyTorch
  • Convolutional filters, padding, and stride.
  • Layer types: convolutional, pooling, flattening, and fully connected
  • Determining shape of the weight matrix for each layer
  • Determining shape of the output for each layer
  • Determining number of parameters
Autoencoders
  • Encoder and decoder structure
  • Latent space representation
  • Variational autoencoders (VAEs)
  • Reparameterization trick
  • Implementation and training
Transformers
  • Multi-Head attention
  • Feed forward network structure
  • Queries, Keys, and Values
  • Positional Encoding
  • Masked multi- head attention
Stable Diffusion
  • Overview of diffusion models: Forward and reverse diffusion processes.
  • Adding and gradually removing Gaussian noise.
  • Multi variant Gaussian distributions
  • U-Net architecture
 
Coding Component
 
  • The exam includes coding questions requiring you to write or complete Python code (using NumPy or PyTorch).
  • Questions will be based on lecture material and assignments.
  • Some problems may closely resemble or be adapted from assignments. 
 
 
 
Top