gradient descent

3Blue1Brown's second neural network chapter explains cost functions and gradient descent, showing how a network learns by minimizing error during training.

A short cartoon explaining how language models like ChatGPT work, from autocomplete and conditional probabilities to neural networks, gradient descent and backpropagation. It shows how networks act as universal function approximators.

Tom 7 (suckerpinch) delivers a long, comedic digression on gradient descent and floating-point imprecision. He shows how rounding behavior can supply the nonlinearity needed to build a neural network.

3Blue1Brown's first chapter on deep learning explains what neurons and layers are and the math underlying a basic neural network, using handwritten digit recognition with visual animations.

3Blue1Brown's third deep learning chapter gives a visual, intuitive account of backpropagation and what happens to a neural network as it learns. A follow-up video covers the underlying calculus.

3Blue1Brown's chapter 4 on deep learning gives a more formal, symbol-based account of backpropagation calculus, connecting the intuition from the previous chapter to the math found in textbooks and code.

MinutePhysics explains that computers store brightness nonlinearly, so blurring and blending colors in raw RGB gives wrong, dark results. It shows why mixed colors, such as in gradients, look off.

NASA JPL engineers explain the final minutes of the Curiosity rover's entry, descent and landing on Mars, which combined aerobraking, a parachute and a skycrane. It lays out the tight sequence of events that all had to work automatically.

NASA's animated explanation of the Curiosity rover's Mars landing, which relied on a complex automated sequence taking about seven minutes. It features NASA engineers discussing the risks before the 2012 landing.

Technology Connections experiments with filtering white and colored LEDs to make holiday lights look less garish. He explains why narrowband LED strings look harsh compared with incandescent bulbs.

Christian Picciolini tells how he became a white supremacist at 14 and led America's first neo-Nazi skinhead gang, then how he left. He shares a counterintuitive approach to countering hate and co-founded Life After Hate.

Technology Connections examines why light bulbs have the lifespans they do, including the real history of the Phoebus cartel. He shows that bulb life trades off against brightness and efficiency rather than being simple planned obsolescence.

A short Harvard Online animated lesson on how the mitochondrial electron transport chain works. It shows proton pumping and the machinery that drives ATP synthesis.

Footage from a camera on a Space Shuttle solid rocket booster from liftoff through descent, with audio enhanced by Skywalker Sound from the original camera microphones. The sound is real, not foley.

The first of a short series by Welch Labs on building a complete neural network in Python. It introduces the data and the architecture of a simple network before later parts cover forward propagation, gradient descent, and backpropagation.

A two-hour driver's-eye cab ride along the Bernina railway from St. Moritz in Switzerland to Tirano in Italy. The UNESCO-listed line crosses the Alps at high altitude with steep adhesion gradients.

Sped-up onboard camera footage of the Falcon 9 first stage landing on a drone ship during the THAICOM 8 mission on May 27, 2016. It shows the descent from grid fin deployment to touchdown.

PBS Space Time explores whether time causes gravity, explaining that gravity does not truly warp the flow of time but arises from time gradients. It revisits how gravitational time dilation works in general relativity.

A GoPro-recorded flight of an RC glider setting a 548 mph dynamic soaring speed record at Parker Mountain, California, using only wind gradients.

SpaceX's official footage of the CRS-6 Falcon 9 first stage landing attempt, in high resolution and slow motion. It shows the descent and touchdown on the drone ship.