PyTorch Workshop

6 hours of instruction

Introduces participants to the foundations of Deep Learning through PyTorch. Participants practice constructing neural networks of various levels of complexity to connect the core ideas to their realization in practical applications (e.g., image processing, natural language processing, etc.).

PREREQUISITES

Participants should be comfortable using the Python language and, in particular, working with standard Python tools for data analysis (notably NumPy, Pandas, and Jupyter). No prior experience using libraries for machine/deep learning (e.g., Scikit-Learn, PyTorch, TensorFlow, etc.) is expected. Some prior exposure to calculus and linear algebra is helpful but not mandatory.

LEARNING OBJECTIVES

1. ​​Explain the relevance of deep learning in application contexts
2. Identify & explain the role of various components of neural network architectures (e.g., layers, neurons/units, activation functions, weights/biases)
3. Implement simple illustrative examples of forward propagation & back-propagation in Python with PyTorch
4. Define descriptive examples of neural networks with various choices of activation functions, loss functions, optimizers, & network architectures
5. Construct working examples of feed-forward and convolutional neural networks of varying depth & complexity using PyTorch