Index

A

action, [1] activation

activation function acyclic auto-differentiation

B

back-propagation, [1] backup

base case behavioral interface Breadth First Search

C

cache, [1] classification binary multi-class collections colocation combine

complexity computation graph conquer convex combination cost cross-entropy cyclic

D

dataclasses decorator deep learning, [1] auto-differentiation back-propagation cost fit, [1] gradient, [1], [2] layer loss, [1] mean square error non-linearity stochastic gradient descent test test set topographic order train set validation set

dependencies Depth First Search Dijkstra directed Directed Acyclic Graph discount divide dynamic programming dynamic programming shortest path

E

edge environment

episode epsilon-greedy exploration

F

fixture forward / backward sweep Forward Neural Network back-propagation layer

function scope

G

generalization generator git github gradient, [1], [2], [3] back-propagation gradient descent, [1] grading criteria graph acyclic Breadth First Search cyclic Depth First Search Dijkstra directed Directed Acyclic Graph edge node shortest path topological order traversal undirected Uniform Cost Search weight

H

higher order derivative

hyper-parameters

I

ide	independent subproblems itertools

L

label layer, [1]

learning rate local repository loss, [1], [2]

M

Markov Decision Process action discount game magic tram optimum policy policy policy evaluation quality reward state swamp isle transition value value iteration, [1]

mean square error memoization model parameters biases weights model-based Monte-Carlo model-free Monte-Carlo

N

neural network

node

O

one-hot encoding oop callable class attributes iterable iterator

optimum policy, [1] over-fitting, [1] overlapping subproblems

P

package PINN loss residual vectorized auto-differentiation pip policy

policy evaluation profiling deterministic memory probabilistic project structure protocol python built-in types

Q

Q-learning

quality

R

recursion recursive case regression regularization, [1] reinforcement learning environment episode epsilon-greedy exploration game learning rate magic tram model-based Monte-Carlo model-free Monte-Carlo optimum policy Q-learning sarsa swamp isle temporal difference utility

remote repository reverse differentiation reward

S

sarsa schedule state, [1] state model action cost dynamic programming magic tram minimum path swamp isle uniform cost search

stochastic gradient descent supervised learning

T

temporal difference test testing assertion fixture parametrization

topographic order topological order training transition traversal type hints abstract type

U

undirected Uniform Cost Search

utility uv

V

validation value value iteration

vectorized auto-differentiation version control vscode

W

weight