The Full Monte Carlo
Plain Monte Carlo integration wastes samples in regions where the integrand is small. Importance sampling fixes this by doing this instead.
What is sampling from a proposal distribution that is concentrated where the integrand is large?
This is the essential thing you need in supervised learning that you do not need in reinforcement learning.
What is labeled training data (input-output pairs with known correct answers)?
SOR stands for this.
What is Successive Over-Relaxation?
In the Linear Congruential Generator
r_{i+1} = ar_i+c mod M
this is the maximum possible period of the sequence.
What is M?
In supervised learning, this is the difference between a classification problem and a regression problem.
What is discrete (categorical) outputs vs. continuous outputs?
In the iterative solution of Laplace's equation, the Gauss-Seidel method differs from the Jacobi method in this specific way.
What is that updated values are used immediately within the same sweep, rather than waiting until the next iteration?
In a Metropolis simulation, this is the term for the initial transient period before the chain has "forgotten" its starting point, during which samples must be discarded.
What is burn-in (or thermalization)?
In a neural network, this algorithm efficiently computes how each weight should be updated by propagating gradients backward through the network.
What is backpropagation?
In the finite difference solution of Laplace's equation, the value at each interior grid point equals this function of its neighbors.
What is the average of the four surrounding grid point values?
This is the reason Monte Carlo integration is preferred over grid-based quadrature for integrals in 10 or more dimensions.
What is that its error is dimension-independent (it avoids the curse of dimensionality)?
n reinforcement learning, the agent receives this signal from the environment in place of labeled targets, which it uses to improve its policy over time.
What is a reward?
The notebook simulates a wavepacket hitting a potential barrier. When the barrier height is finite, part of the packet reflects and part transmits. The transmission of a particle through a classically forbidden barrier is this quantum phenomenon.
What is quantum tunneling?