Application track
Slide 1 is the company list. Slides 2 to 4 are the concepts, each linked to a live lesson.
Oncology data and AI diagnostics
Matrices of patient features; linear models for risk scores
Pathology image recognition
Convolution on tissue slides (same idea as fruit edges)
ER imaging triage
Classifiers that rank scans, like apple-or-orange scoring
Drug discovery from cell images
Huge image matrices + learned weights
Handheld ultrasound + AI
Pixels as numbers, then convolution for structure
A tiny window slides over an image and multiplies neighbors. That is how edges, vessels, and tumors get outlined.
Try it on fruit edgesA U-Net is a stack of convolutions that labels every pixel (organ vs background). It starts from the same RGB grid as Lesson 1.
See pixels as numbersPhysics-Informed Neural Networks add calculus: the model must also obey a derivative equation, not just fit labels.
Open the Math Lab