A gentle introduction to Deep Learning
summary
Begins with simple math and builds a working image classifier that distinguishes pizza from hot dog using about 883 training images, trained roughly 25 epochs and hitting ~73% accuracy in a quick demo. Uses Python, PyTorch and Jupyter for development, saves the model as ONNX and deploys it to Azure Machine Learning then to Azure Container Instances. Covers concrete issues—gradient descent, overfitting and dropout—and shows real failure modes and ethical problems like biased medical data.
AI-written summary · about
chapters (10)
- 0:00Introduction to AI concepts
- 4:02Fundamentals of learning
- 10:49Neural network architecture
- 18:57Layers and activation functions
- 22:33Training process example
- 27:02Learning methodologies
- 29:19Error reduction and training
- 37:38Technical demo: Pizza or hot dog
- 50:04Overfitting and pitfalls
- 55:10Ethics, resources, and wrap-up
related talks
59mA Friendly Introduction to Deep Learning for Computer Vision
32mYou Can Do Deep Learning!
23mBuilding high-quality computer vision models using only a few examples
35mIs it Food? An Introduction to Machine Learning
01h 05mDeep Learning with PyTorch
01h 01mA practical guide to deep learning
01h 01mA practical guide to deep learning
01h 29mComputer Vision: An (Un?)Expected Journey, with Keras and Tensorflow