Causality

Can time-traveling cause a future event?

Imagine you're watching a movie where a character flips a coin. You see the coin flip (cause) and then it lands (effect). You wonder if the coin flip caused the landing.

In time-series data, we want to know if a past event (like flipping a coin) caused a future event (like the coin landing). The do-calculus helps us figure this out by looking at the sequence of events over time.

Example

Coin flip (time 1) → Coin lands (time 2).

Remember this

The do-calculus helps us understand if one event in time caused another, just like knowing if the coin flip caused the coin to land.

Text adapted from Wikipedia, licensed under CC BY-SA 4.0.

Related concepts

the do-calculus does: computes interventional probabilities from observational data

Can we predict the future from the past?

the reverse process learns: p_θ(x_{t-1}|x_t)

Can we trace back the roots of life?

Likelihood function

Why can't we just guess probabilities correctly all the time?

Dirac delta function

How can we predict random events with certainty?

Physics-informed neural networks

Neural ODEs model continuous-time dynamics with a neural network as the derivative

Write the Bellman equation for reinforcement learning

Predicting tomorrow's weather with today's clues

Swipe through more Machine Learning concepts

Open Pocket Polymath