COSC 410A/LA: Midterm Exams
Here are a list of potential topics for each exam. You should make sure read the textbook, review slides/notes/handouts, and take advantage of the other resources that are available to you on the website and elsewhere.
Midterm Exam 1
- For a problem, identify
- What type of learning you would apply (supervised vs. unsupervised)?
- What model is appropriate (linear regression, softmax regression, logistic regression, SVMs, decision trees, k-nearest neighbors, k-means clustering) and why?
- What type of evaluation makes sense (MAE, MSE, accuracy, precision, recall, f1) and why?
- Identify a good cost function
- Apply a model to determine an output
- Calculate the cost associated with a prediction
- Describe gradient descent with reference to the concepts on a figure (including local vs. global minima)
- The linear regression formula and the trade-offs of closed-form vs. gradient descent optimization
- Different treatments of data with gradient descent (batched, mini-batch, and stochastic) and their trade-offs
- Calculate precision, recall, and f1
- Describe, and graphically demonstrate, k-nearest neighbors
- Describe, and graphically demonstrate, k-means clustering
- Apply the CART algorithm
- Describe the application of the ML pipeline to a problem
- K-fold validation
Midterm Exam 2
- For a problem, identify
- What type of neural network is appropriate (RNN, CNN, Transformer) and why?
- What type of architecture is needed (encoder, decoder, encoder-decoder)?
- What evaluation metric makes sense and why?
- Apply neural network mechanisms, recurrence, kernel, pooling, and attention mechanism, to a concrete example
- Describe different approaches to recommender systems, their strengths/weaknesses, and good uses cases for each
- Calculate recommendation from a recommendation system for a small example
- Describe different autoencoder architectures and their purpose
- Concrete apply q-value iteration and q-learning to small examples
- Build a MDP for a problem description
- Describe hyperparameter tuning
- Understand a q-value matrix
- Use convolution to calculate a specified operation