Cheat SheetsMath for AIHigh-School Refresher

High-School Refresher — Cheat Sheet

Math for AI · 3 topics. Download the PDF or the Instagram carousel and share it.

Cheat Sheet · AiCanCode.org
High-School Refresher
Math for AI3 topicsQuick revision reference
1

Do You Need a PhD? The AI-Math Mindset

You do not need a PhD in mathematics to be a strong AI engineer — you need a solid, intuition-first grasp of a handful of topics and the ability to connect each one to what a model actually does.

  • No PhD required — intuition + code + knowing where each topic is used is enough to start
  • Top priority: Linear Algebra, Calculus, Probability & Statistics
  • Libraries do the heavy computation; you supply the understanding
  • Always ask "what is this measuring/doing?" before memorising notation
The map: math topic → where it shows up in AI
Linear Algebra   -> data as vectors/matrices, embeddings, every neural-net layer
Calculus         -> how a model learns (gradients, backpropagation)
Probability      -> reasoning under uncertainty, LLM next-token sampling
Statistics       -> evaluating models, understanding data
Optimization     -> training efficiently (gradient descent, Adam)
Information Theory-> loss functions (cross-entropy), comparing distributions
Discrete Math    -> algorithms, knowledge graphs, search
Graph Theory     -> graph neural nets, recommendations
2

Functions, Graphs & Coordinate Geometry

A function is a machine that maps inputs to outputs; a graph is its picture — and a neural network is just a very large, learnable function.

  • A function maps inputs to outputs; its graph is that mapping drawn out
  • y = wx + b: slope=weight, intercept=bias — the core of a linear layer
  • Non-linear functions (ReLU, sigmoid) let networks fit curved data
  • Reading a graph = predicting a function's behaviour without computing every point
y = wx + b — one line, two learnable numbers
import numpy as np

w, b = 2.0, 1.0                 # weight (slope), bias (intercept)
x = np.linspace(-3, 3, 7)       # inputs: -3, -2, ... 3
y = w * x + b                   # the line / a linear neuron
print(y)   # [-5. -3. -1.  1.  3.  5.  7.]

# Change w -> steeper line; change b -> whole line moves up/down.
# Training a model = searching for the w and b that fit the data.
3

Exponents, Logarithms & Summation Notation

Exponentials grow explosively, logarithms tame that growth back down, and the Σ (sigma) symbol is just a for-loop — three notations you will meet on every page of AI math.

  • e^x grows fast and stays positive; log is its inverse and compresses scale
  • ML uses log-probabilities to avoid underflow and to shape loss functions
  • Σ (sigma) means "iterate and add" — a for-loop in disguise
  • log turns products into sums, which is why log-likelihood is everywhere
log turns "multiply many small numbers" into "add"
import numpy as np

p = np.array([0.9, 0.2, 0.01])   # three probabilities
print(np.log(p))                 # [-0.105 -1.609 -4.605]  (more negative = more surprising)

# multiplying probabilities vs adding their logs (same ranking, safe math):
print(np.prod(p))                # 0.0018
print(np.exp(np.sum(np.log(p)))) # 0.0018  -> log-sum then exp recovers it
Learn this free with Aria, your AI tutor → AiCanCode.org/learn/math-for-ai