Index Ai Academy
Foundation · Vectors and matrices

Matrix multiplication

Row meets column, the size rule, order matters, and a batch through a layer: X W + b.

About 45 minutes · Free lesson

What it covers6

Why it matters

A matrix times a vector does several dot products for one vector. A matrix times a matrix does them for many vectors at once: every row of the first meets every column of the second. That one operation is the workhorse of modern AI. A layer of a neural network applied to a whole batch of examples is one matrix multiplication, and the attention inside a transformer is a few more. GPUs are, more or less, machines built to do it fast. This lesson builds it from shopping bills, works it by hand with the size rule as a guide, writes it as three loops and then as one @, and finishes by pushing four pictures through a layer in a single line, in NumPy and in PyTorch.

What you'll be able to do

  • Multiply two matrices: entry (i, j) is row i of the first dotted with column j of the second
  • Use the size rule: inner sizes must match, and the outer sizes give the answer's size
  • Compute A @ B with a triple loop and with @, time the two, and know that A @ B and B @ A usually differ
  • Run a batch of pictures through a layer at once, Z = X W + b, and get the same scores from PyTorch's nn.Linear
Quick check
Have a guess before you start. A 2 × 3 matrix times a 3 × 2 matrix. How many numbers are in the answer?
  1. 4
  2. 6
  3. 9
  4. 12
Show the answer

4

The answer is 2 × 2: one entry for each row of the first (2) and each column of the second (2).

The math: every row meets every column

Purpose. Many bills at once. Three hikers each buy some energy bars, apples and nuts, recorded in A, one row per hiker. Two shops price the three items, recorded in B, one column per shop:

A=(213041520)B=(1.52.00.50.43.02.5)A = \begin{pmatrix} 2 & 1 & 3 \\ 0 & 4 & 1 \\ 5 & 2 & 0 \end{pmatrix} \qquad B = \begin{pmatrix} 1.5 & 2.0 \\ 0.5 & 0.4 \\ 3.0 & 2.5 \end{pmatrix}

Hiker 1's bill at shop 1 is a dot product, row 1 of A with column 1 of B: 2 × 1.5 + 1 × 0.5 + 3 × 3 = 12.5. Do that for every hiker and every shop and you have six bills in a 3 × 2 table. That table is the matrix product AB.

Quick check
From lesson 55: a 3 × 3 matrix times a vector of 3. How many numbers come out?
  1. 3
  2. 6
  3. 9
  4. 1
Show the answer

3

One dot product per row. A matrix product is that, repeated for every column of the second matrix.

The picture. Step through the six bills. The first step checks the sizes; each later step lights a row of A and a column of B, pairs them up, and drops their dot product into the answer. The drawing lays the three tables out so they meet: B sits above the answer and A to its left, with an empty corner between them (the arrows "B →" and "A ↓" point across it), and each bill shows ? until it's made. Row i of A runs straight across to row i of the answer, and column j of B runs straight down to its column j, so the cell where they cross is the bill they make.

Try it now

Then set matrix B to two lines, 1 0 and 0 1, and press Use these numbers: the first step refuses, because the inner sizes no longer match.

Symbols and formula. For A of size m × n and B of size n × p:

Cij=∑k=1nAik BkjC_{ij} = \sum_{k=1}^{n} A_{ik} \, B_{kj}

In plain words: row i of A dotted with column j of B. Some people say "over and down": go across the row, down the column, multiplying pairs as you go.

The size rule. Write the sizes side by side, (m × n)(n × p). The inner sizes must match, because each row of A has n entries and each column of B has n entries to pair with them. The outer sizes give the answer: m × p. (3 × 3)(3 × 2) gives 3 × 2; (2 × 3)(2 × 3) can't be done.

The size rule as a picture. Think of a matrix product as a machine that a table travels through. The arrow into the machine carries the first table's size; the machine holds the second table; the arrow out carries the answer's size. Step through three uses of it.

Quick check
Can a 2 × 3 matrix multiply another 2 × 3 matrix?
  1. Yes: the two sizes are exactly the same
  2. Yes, giving a 2 × 3 answer
  3. Yes, giving a 3 × 3 answer
  4. No: the inner sizes, 3 and 2, don't match
Show the answer

No: the inner sizes, 3 and 2, don't match

(2 × 3)(2 × 3): the inner sizes are 3 and 2. A row of 3 has nothing to pair with in a column of 2.

Order matters. With numbers, 3 × 4 = 4 × 3. With matrices, AB and BA are usually different, and often only one of them can even be formed. For P = [[1, 2], [3, 4]] and Q = [[0, 1], [1, 0]], PQ = [[2, 1], [4, 3]] swaps P's columns, while QP = [[3, 4], [1, 2]] swaps its rows. Always multiply in the order the problem gives.

A named mistake: multiplying matching cells. It's tempting to multiply entry by entry, as adding does. That's a different operation with a different answer, and it only works when both matrices are the same size.

In your own words
In your own words: why must the inner sizes match in a matrix product?

Write a sentence or two, then compare it with a good answer in the lesson.

In Python: three loops, then @

Purpose. Six bills by hand is fine; six million is not.

The formula has three letters, i, j and k, so a direct translation has three loops:

import numpy as np

A = np.array([[2, 1, 3], [0, 4, 1], [5, 2, 0]])
B = np.array([[1.5, 2.0], [0.5, 0.4], [3.0, 2.5]])
C = np.zeros((3, 2))
for i in range(3):
    for j in range(2):
        for k in range(3):
            C[i, j] = C[i, j] + A[i, k] * B[k, j]
print(C.tolist())
print((A @ B).tolist(), np.allclose(C, A @ B))
It runs right here in your browser, with nothing to install.

@ does the same work in one character, and much faster. np.allclose checks two arrays are equal up to tiny float differences, which is the right test for decimals.

Code flow. Step through the three loops for the first two hikers, a line at a time. i picks a hiker, j a shop, and k runs along the hiker's row and down the shop's column together, adding one pair's product to C[i, j] on each pass. That's the math part's row meeting a column, one product at a time.

How much faster is @? Time both on bigger tables. time.perf_counter() reads a stopwatch in seconds, so the difference between two readings is how long the lines between them took. np.ones makes tables full of 1s, and @ is timed over 100 runs, because one run is too quick to measure well.

import time
import numpy as np

X = np.ones((40, 30))
W = np.ones((30, 20))
start = time.perf_counter()
Z = np.zeros((40, 20))
for i in range(40):
    for j in range(20):
        for k in range(30):
            Z[i, j] = Z[i, j] + X[i, k] * W[k, j]
loop_time = time.perf_counter() - start
start = time.perf_counter()
for rep in range(100):
    Z2 = X @ W
at_time = (time.perf_counter() - start) / 100
print(np.allclose(Z, Z2), loop_time > at_time)
print("loop:", round(loop_time * 1000, 2), "ms   @:", round(at_time * 1000, 4), "ms")

The loop works through 40 × 20 × 30 = 24,000 products one at a time in Python; @ hands the whole job to NumPy's compiled code. Your times change from run to run, but the loop is always far slower. Swapping a Python loop for one array operation like this is called vectorization, and it's why real code never multiplies matrices by loop.

Print the shapes. The size rule in code is a rule about .shape:

import numpy as np

A = np.array([[2, 1, 3], [0, 4, 1], [5, 2, 0]])
B = np.array([[1.5, 2.0], [0.5, 0.4], [3.0, 2.5]])
print(A.shape, B.shape, (A @ B).shape)
print(A.shape[1] == B.shape[0])

When the inner sizes differ, @ stops with an error that mentions a "mismatch in its core dimension". Read it as "the inner sizes don't match", and print both shapes.

A named mistake: * where @ was meant. On two matrices of the same shape, A * B runs happily and multiplies matching cells. It's not a matrix product, and it gives different numbers. A product of a 2 × 2 and a 2 × 2 is a 2 × 2 either way, so the shape won't warn you; only the values will.

Worked example

Worked example: @ against *, and the order

Multiply two 2 × 2 matrices both ways round with @, and compare with *.

import numpy as np

P = np.array([[1, 2], [3, 4]])
Q = np.array([[0, 1], [1, 0]])
print((P @ Q).tolist())
print((Q @ P).tolist())
print((P * Q).tolist())
In the lesson you can change this and run it.

P @ Q swaps P's two columns, and Q @ P swaps its two rows: the order changes the answer. P * Q multiplies matching cells, keeping only the entries where Q has a 1, which is neither product. Q is a "swap" matrix; lesson 58 shows that every matrix is a movement like this, and multiplying them chains the movements.

Quick check
What does this print?
import numpy as np

A = np.array([[1, 2, 3],
              [4, 5, 6]])
B = np.array([[1, 0],
              [0, 1],
              [2, 1]])
print((A @ B).shape, (A @ B).tolist()[1])
  1. (2, 3) [4, 0, 12]
  2. (2, 2) [7, 5]
  3. (2, 2) [16, 11]
  4. (3, 3) [16, 11]
Show the answer

(2, 2) [16, 11]

(2 × 3)(3 × 2) gives 2 × 2. Row 2 of A with each column of B: 4 + 0 + 12 = 16 and 0 + 5 + 6 = 11.

Quick check
A has shape (5, 8) and B has shape (8, 3). What is the shape of A @ B?
  1. (8, 8)
  2. (3, 5)
  3. (5, 8)
  4. (5, 3)
Show the answer

(5, 3)

The inner 8s match and disappear; the outer sizes, 5 and 3, are left.

Drill

The size rule as code

Write can_multiply(a_shape, b_shape), True when the inner sizes match, and result_shape(a_shape, b_shape), the answer's shape as a tuple, or None when they can't be multiplied. Shapes are tuples like (3, 2).

def can_multiply(a_shape, b_shape):
    return a_shape[0] == b_shape[1]

def result_shape(a_shape, b_shape):
    return (a_shape[1], b_shape[0])

print(can_multiply((3, 2), (2, 4)), result_shape((3, 2), (2, 4)))
In the lesson, your code is checked as you write it.
In your own words
Why doesn't the shape catch a * written where @ was meant, for two square matrices?

Write a sentence or two, then compare it with a good answer in the lesson.

In AI and ML: a layer is one matrix multiplication

Purpose. Run a whole batch of examples through a layer at once.

Stack four pictures, each unrolled to 25 pixels, as the rows of X (4 × 25). Give a layer with three outputs a weight matrix W of 25 × 3: one column of 25 weights per output, the transpose of lesson 55's layout. Then

Z=XW+bZ = XW + \mathbf{b}

In plain words: every picture's row meets every output's column, and each output's bias is added to its column. Z is 4 × 3: one row of three scores per picture.

Here it is on a tiny batch of two examples with three features each, so every step is visible. The picture draws the layer as a network: the current example's numbers fill the circles on the left, each column of W rides the lines into one output on the right, and each output starts at its bias. The examples pass through one at a time.

Now the real sizes. The weights here are the three sign templates from lesson 55, one per column:

import numpy as np

cross = np.array([[1,0,0,0,1],[0,1,0,1,0],[0,0,1,0,0],[0,1,0,1,0],[1,0,0,0,1]])
arrow = np.array([[0,0,1,0,0],[0,1,1,1,0],[1,0,1,0,1],[0,0,1,0,0],[0,0,1,0,0]])
box = np.array([[1,1,1,1,1],[1,0,0,0,1],[1,0,0,0,1],[1,0,0,0,1],[1,1,1,1,1]])
W = np.array([cross.flatten(), arrow.flatten(), box.flatten()]).T
X = np.array([arrow.flatten(), box.flatten(), cross.flatten(), arrow.flatten()])
b = np.array([0, 0, 0])
Z = X @ W + b
print(X.shape, W.shape, Z.shape)
print(Z.tolist())

Each row of Z is one picture's three scores, and the biggest in each row names its sign: arrow, box, cross, arrow.

Quick check
What does this print?
import numpy as np

X = np.array([[1, 2], [0, 1], [3, 0]])
W = np.array([[1, 0, 1], [0, 1, 1]])
print((X @ W).shape, (X @ W).tolist()[2])
  1. (2, 2) [3, 0, 3]
  2. (3, 3) [3, 0, 3]
  3. (3, 2) [3, 0]
  4. (3, 3) [1, 2, 3]
Show the answer

(3, 3) [3, 0, 3]

(3 × 2)(2 × 3) gives 3 × 3. Row 3 of X, (3, 0), meets each column of W: 3, 0 and 3.

In PyTorch. PyTorch's layer, nn.Linear(25, 3), stores its weights the way lesson 55 did, as (out, in): one row of 25 per output, 3 × 25. That's the transpose of this part's W. So for a batch whose examples are rows, the layer computes X @ layer.weight.T + layer.bias: the same Z = XW + b, with W written as the transpose of what it stores. The cell loads the three templates as the layer's rows, works the formula by hand, and checks it against layer(X). torch.stack stacks tensors as rows and .float() makes them decimals, as in lesson 55.

import torch
import torch.nn as nn

cross = torch.tensor([[1,0,0,0,1],[0,1,0,1,0],[0,0,1,0,0],[0,1,0,1,0],[1,0,0,0,1]]).float()
arrow = torch.tensor([[0,0,1,0,0],[0,1,1,1,0],[1,0,1,0,1],[0,0,1,0,0],[0,0,1,0,0]]).float()
box = torch.tensor([[1,1,1,1,1],[1,0,0,0,1],[1,0,0,0,1],[1,0,0,0,1],[1,1,1,1,1]]).float()
X = torch.stack([arrow.flatten(), box.flatten(), cross.flatten(), arrow.flatten()])
layer = nn.Linear(25, 3)
with torch.no_grad():
    layer.weight.copy_(torch.stack([cross.flatten(), arrow.flatten(), box.flatten()]))
    layer.bias.fill_(0)
Z = X @ layer.weight.T + layer.bias
print(layer.weight.shape, Z.shape)
print(Z.tolist())
print(torch.equal(Z, layer(X)))

Z holds the scores the NumPy cell printed, now as decimals: arrow, box, cross, arrow again. torch.equal says the formula and layer(X) agree exactly. That one line, X @ W.T + b on a whole batch, is what every nn.Linear in a real model computes.

Quick check
The layer above is changed to have 10 outputs instead of 3. What shape must this part's W be, for X @ W?
  1. (10, 25)
  2. (4, 10)
  3. (25, 10)
  4. (25, 3)
Show the answer

(25, 10)

W needs one row per input pixel (25) and one column per output (10). PyTorch's nn.Linear(25, 10) would store its transpose, (10, 25).

Drill

Score a new batch

Three smudged pictures are in Xb, one per row. Set Z to the layer's scores, X @ W + b, and labels to the position of the largest score in each row, as a list.

import numpy as np

cross = np.array([[1,0,0,0,1],[0,1,0,1,0],[0,0,1,0,0],[0,1,0,1,0],[1,0,0,0,1]])
arrow = np.array([[0,0,1,0,0],[0,1,1,1,0],[1,0,1,0,1],[0,0,1,0,0],[0,0,1,0,0]])
box = np.array([[1,1,1,1,1],[1,0,0,0,1],[1,0,0,0,1],[1,0,0,0,1],[1,1,1,1,1]])
W = np.array([cross.flatten(), arrow.flatten(), box.flatten()]).T
b = np.array([0, 0, 0])
a2 = arrow.copy()
a2[0, 0] = 1
b2 = box.copy()
b2[2, 2] = 1
Xb = np.array([a2.flatten(), b2.flatten(), cross.flatten()])
Z = Xb * W + b
labels = [0, 0, 0]
print(Z, labels)
In the lesson, your code is checked as you write it.
In your own words
Why is it useful that a whole batch goes through a layer as one matrix product, instead of one picture at a time?

Write a sentence or two, then compare it with a good answer in the lesson.

Goes deeper in

Practice what you just read

  • 6 code examples you can edit and run
  • a practice exercise that checks your code
  • a 5-question quiz
  • 9 review cards for spaced repetition

Try a quiz question

What is [[1, 2], [0, 1]] times [[3, 0], [1, 2]]?

  1. [[5, 4], [1, 2]]
  2. [[3, 0], [0, 2]]
  3. [[3, 2], [1, 3]]
  4. [[4, 2], [1, 3]]

Answer it at the end of the lesson to see if you're right, and why.

Before this lesson

Keep learning

Where this lesson leads.

575 lessons in thirteen phases. 176 are free without an account, and a free account opens the rest. See every lesson.

Foundation · about 96 hours of study

Foundation math

From whole numbers to the math inside a neural network: each idea by hand, in Python, and in AI 118 lessons All free
  1. Numbers and place value FreeWhat a digit's place is worth, the number line, and Python as a calculator that never tires.36 min
  2. Multiplying and dividing: groups FreeEqual groups, quotient and remainder, and cutting training data into batches.35 min
  3. Order of operations, and names for numbers FreeBrackets, powers, then multiply and divide, then add and subtract; variables hold numbers.36 min
  4. Numbers below zero FreeThe number line through zero, sign rules, distance from zero, and an error's direction and size.37 min
  5. Fractions: equal parts, equal fractions, adding FreePieces of a whole, equal fractions, and re-cutting to add; accuracy as a fraction.45 min
  6. Multiplying and dividing fractions FreeA share of a share, dividing by flipping, and why chances in a row get small.35 min
  7. Decimals and rounding FreeFractions in place value, rounding, and why 0.1 + 0.2 isn't exactly 0.3.36 min
  8. Percentages FreePer hundred, percent change, percentage points, and the accuracy and error rates of models.35 min
  9. Ratios, rates and proportion FreeSharing in a ratio, rates per unit, scaling, and splitting data 80/10/10.35 min
  10. Powers and roots FreeRepeated multiplying, the laws of exponents, square roots, and squared error.43 min
  11. Very big and very small numbers FreePowers of ten, e-notation like 3e-4, orders of magnitude, and a model's size in memory.36 min
  12. Bits, bytes and floats: how numbers are stored FreeBase 2, what n bits can hold, a float as sign, power and digits, and a tensor's memory.35 min
  13. Review: find the slip (numbers) FreeBroken calculations to fix, mixed practice from the whole module, and its formula sheet.34 min
  14. Letters for numbers FreeVariables, terms and like terms, substitution, and a model's formula as a Python function.41 min
  15. Expanding and factoring FreeBrackets as areas, pulling out common factors, and a squared error expanded into a bowl.41 min
  16. Solving equations: keep the balance FreeThe same move on both sides, undoing steps in reverse, and running a model's formula backward.41 min
  17. Rearranging a formula FreeMaking any letter the subject, round-trip checks, and planning a training run.36 min
  18. Inequalities and true-or-false FreeRanges, flipping the sign, if and elif, and the thresholds, clips and ReLU inside a model.42 min
  19. Functions: a rule from input to output Freef(x) notation, tables and graphs, domain and range, loops over lists, and loss as a function.42 min
  20. Chaining and undoing functions Freeg(f(x)), inverses that undo, scaling and unscaling, and a network as a chain.42 min
  21. Straight lines: slope and intercept FreeRise over run, the line through two points, and a model's weight and bias as slope and intercept.42 min
  22. Two equations, two unknowns FreeSubstitution and elimination, where lines cross, and why models are fitted, not solved.42 min
  23. Quadratics: bowls and their lowest point FreeRoots by factoring and by formula, the discriminant, the vertex, and the best weight of a model.42 min
  24. Exponentials: growth and decay FreeThe same factor each step, the number e, and decaying learning rates.43 min
  25. Logarithms: the power question FreeTo what power?, the laws of logs, and adding log-probabilities instead of multiplying.45 min
  26. Sequences and Σ FreeArithmetic and geometric sequences, Σ as a loop that adds, and mean squared error.40 min
  27. Weighted and moving averages FreeWeights that add to 1, a window average, the exponential moving average, and smoothing a loss curve.35 min
  28. Sets, and reading the symbols FreeSets, union and intersection, ℝⁿ and f : ℝⁿ → ℝᵐ, and reading a paper's first equation.41 min
  29. Review: find the slip (algebra) FreeBroken algebra and code to fix, mixed practice from the whole module, and its formula sheet.34 min
  30. Transposing step by step: undo in reverse order FreeA formula's layers, the opposite of each, peeling from the outside in, conditions, and round-trip checks.35 min
  31. Formulas with fractions: clear them first FreeClearing denominators, cross-multiplying, a letter in the bottom, and reciprocal formulas like 1/R = 1/R₁ + 1/R₂.35 min
  32. When the letter appears twice: collect and factor FreeGathering the letter on one side, factoring it out, dividing by the bracket, and rearranging precision and F1.35 min
  33. Squares, roots and powers FreeSquare roots with ±, squaring to undo a root, cube and nth roots, and the √d inside attention.40 min
  34. When the letter is squared and plain: quadratics FreeRearranging into ax² + bx + c = 0, the formula, completing the square, and choosing the root that fits.35 min
  35. Letters in the power: exponentials and logarithms FreeTaking logs to bring a power down, exponentiating to free a letter, and the logit that undoes the sigmoid.40 min
  36. Rearranging inequalities: when the sign flips FreeThe four moves on an inequality, negatives and reciprocals that flip it, logs below 1, and sample sizes.35 min
  37. Inverse functions, lines and scaling FreeUndoing a function by making x the subject, domains that allow it, line forms, and unscaling predictions.40 min
  38. Sequences and sums: find n, a, d and r FreeEvery letter of the arithmetic and geometric formulas as the subject, and an average's window from its decay.40 min
  39. Combining formulas: substitute and eliminate FreePutting one formula inside another to remove a letter, and sizing a training run from a time or money budget.35 min
  40. Review: rearrange anything (algebra) FreeBroken rearrangements to fix, mixed practice from the whole algebra block, a strategy chart, and its formula sheet.35 min
  41. Coordinates and distance FreePoints as (x, y), Pythagoras, distance in any number of coordinates, and nearest neighbors.42 min
  42. Angles, degrees and radians FreeMeasuring a turn, converting units, arcs, and the angle between two arrows.36 min
  43. Sine and cosine FreeAcross and up: right triangles, the unit circle, and directions and rotations.42 min
  44. Waves: height, speed and shift FreeAmplitude, period and phase, sampling a wave, and position codes in transformers.37 min
  45. Review: find the slip (shapes and waves) FreeBroken geometry and code to fix, mixed practice from the module, and its formula sheet.29 min
  46. Angles, triangles and waves: rearranging with sin, cos and tan FreeSides and angles from SOH CAH TOA, principal values, arcs, polar coordinates, and a wave's time and period.45 min
  47. Vectors: lists of numbers that point FreeA list, a point and an arrow; components, dimension, NumPy arrays, and data as vectors.42 min
  48. Adding and scaling vectors FreeComponent-wise sums, scalar multiples, mixes, and the weight update w − ηg.43 min
  49. Length and distance: norms FreeL2, L1 and max norms, unit vectors, distance, and the norms inside losses and clipping.42 min
  50. The dot product FreeMultiply in pairs and add; a·a as squared length; a @ b; one neuron as a dot product.43 min
  51. Angle, cosine similarity and projection FreeDirection without length, perpendicular vectors, shadows, and search by meaning.45 min
  52. Perpendicular directions: bases and coordinates FreeOrthonormal directions, coordinates as dot products, rebuilding a vector, and taking a direction out.35 min
  53. Matrices: tables of numbers FreeRows, columns and entries; two-dimensional arrays and slices; a data set as a matrix.41 min
  54. Adding, scaling and transposing matrices FreeEntry-by-entry sums and scalings, the transpose, mean images, and tables turned around.36 min
  55. A matrix times a vector FreeEach row dotted with the vector, the size rule, A @ x, and a layer acting on one example.42 min
  56. Matrix multiplication FreeRow meets column, the size rule, order matters, and a batch through a layer: X W + b.45 min
  57. Identity, diagonal, triangular and symmetric FreeFour patterns and their jobs: rescaling, distance tables, and the causal mask.37 min
  58. Matrices as movements of space FreeStretches, flips, shears and rotations; columns as landing spots; chaining; a layer moving data.42 min
  59. Determinant and inverse FreeThe area factor ad − bc, singular matrices, undoing a matrix, and redundant features.42 min
  60. Solving systems by elimination FreeAugmented matrices, row moves, back-substitution, the three outcomes, and fitting weights.45 min
  61. Independence, span and rank FreeRedundant vectors, span, rank, outer products, and low-rank updates like LoRA.42 min
  62. Eigenvectors: directions that only stretch FreeAv = λv, checking candidates, np.linalg.eig, and the main direction of a data cloud.41 min
  63. SVD and PCA: the important directions first FreeA matrix as pieces by size, the best rank-k copy, compression, and PCA maps of data.45 min
  64. Tensors, shapes and broadcasting FreeAxes and shapes, reshape, axis=, broadcasting, and shapes through a small model.43 min
  65. One-hot vectors and lookup tables: embeddings FreeA one-hot row picks a table's row, a sentence looked up as vectors, and a score for every word.35 min
  66. Sliding dot products: convolution FreeA small pattern slid along a list and over a grid, output sizes, stride and padding, and shared weights.40 min
  67. Review: find the slip (vectors and matrices) FreeShape slips to fix, mixed practice from the module, and its formula sheet with shapes.36 min
  68. Vector equations: rearranging with arrows FreeSolving for a vector or a scale, recovering a gradient from two weight snapshots, and an angle from a dot product.40 min
  69. Matrix equations: multiply by the inverse, on the correct side FreeNo dividing by a matrix: inverses from the left or right, order, transposes, and the normal equations.40 min
  70. Rates of change FreeChange over change, secants, rates from readings with np.diff, and reading a loss curve.41 min
  71. Limits: what a value heads toward FreeApproaching from both sides, holes, corners and jumps, epsilon, and judging activations.37 min
  72. The derivative: slope at a point FreeShrinking secants, f′(x) and dy/dx, Δy ≈ f′Δx, and which way a weight should move.45 min
  73. Derivative rules FreeConstants, powers, multiples, sums, eˣ and ln x, checked numerically, and a squared error's slope.41 min
  74. The product rule and the chain rule FreeProducts and chains of functions, forward then backward, and one neuron's dL/dw.45 min
  75. Lowest points and curvature Freef′ = 0, the second-derivative test, tables, local and global minima, and convex losses.43 min
  76. Standing in for a curve near a point FreeTangent lines, a curvature term, growing error, and why learning rates must be small.37 min
  77. Partial derivatives FreeOne input at a time, ∂, nudging in Python, and the two slopes of a line's loss.41 min
  78. The gradient FreeAll the slopes as a vector, steepest climb and descent, a numerical gradient, and the loss gradient.42 min
  79. Gradient descent FreeStepping against the gradient, the learning rate, a training loop, and fitting a line.45 min
  80. Faster descent: momentum and Adam FreeMomentum and Adam by hand, the bias correction, and three optimizers down one narrow valley.40 min
  81. The chain rule with many paths: backpropagation by hand FreeGraphs, multiply along paths and add across them, gradient checks, and a 2-2-1 network by hand.45 min
  82. Automatic differentiation: how PyTorch finds every slope FreeA tape of steps walked backward, and requires_grad, backward(), .grad, zero_grad() and no_grad().40 min
  83. Jacobians and Hessians FreeSlopes of vector functions, a layer's Jacobian, second partials, and bowls, domes and saddles.42 min
  84. Integrals: adding up thin slices FreeStrip and trapezoid sums, summing undoes differencing, probability as area, and AUC.43 min
  85. Review: find the slip (calculus) FreeCalculus slips to fix, mixed practice from the module, and its formula sheet.36 min
  86. Rearranging in calculus: slopes, steps and constants FreeSetting a slope to zero, Newton's step, the chain rule turned around, constants from a point, and learning rates.45 min
  87. Counting: choices, orders and groups FreeThe counting principle, factorials, ordered picks and groups, and why attention's cost grows with the square of the length.43 min
  88. Probability: from counts to chances FreeEqually likely outcomes, the complement, distributions, seeded simulation, and a next-word model.43 min
  89. Combining events: and, or, independent FreeThe or rule, independence, probability trees, drawing with and without replacement, and chains of steps.42 min
  90. Conditional probability and Bayes' rule FreeConditionals from tables, why P(A | B) isn't P(B | A), Bayes' rule, and precision and recall.45 min
  91. Random variables and distributions FreeUncertain numbers, the probability of a range, sampling by slices, and sampling versus argmax.42 min
  92. Expected value, variance and standard deviation FreeProbability-weighted averages, spread in the data's own units, and a training loss as an expected value.42 min
  93. Bernoulli, binomial, uniform and normal FreeYes-or-no trials, counts of yeses, the 68-95-99.7 rule, z-scores, and an accuracy's error bar.45 min
  94. Two variables at once: covariance and correlation FreeJoint tables and marginals, covariance, correlation, the covariance matrix, and features that carry the same information.45 min
  95. Review: find the slip (probability) FreeProbability slips to fix, mixed practice from the module, and its formula sheet.36 min
  96. Probability formulas: solve for the missing chance FreeConditionals, Bayes' rule for the prior, the or rule, expected values, variance, and z-scores turned around.45 min
  97. Describing data: center, spread and scale FreeMean, median and mode, variance and standard deviation, percentiles, outliers, and standardizing features.42 min
  98. Samples, the law of large numbers and the central limit theorem FreeSamples and populations, the standard error σ/√n, the bell of averages, and error bars on test scores.43 min
  99. Is the difference real? Comparing two models FreePaired differences, a bootstrap interval, what a p-value is and isn't, and why many tries make lucky wins.40 min
  100. Maximum likelihood FreeLikelihoods as products, log-likelihoods as sums, grid search, underflow, and the negative log-likelihood loss.45 min
  101. Linear regression: the best line FreeResiduals, least squares by formula, three roads to one line, the matrix form, and train versus test error.45 min
  102. Logistic regression: a line that answers yes or no FreeThe sigmoid, the log loss, its slope p − y, a training loop, and choosing a threshold.45 min
  103. Scoring a classifier: confusion table, precision, recall, F1 FreeThe four cells, accuracy, precision, recall and F1, and what moving a threshold does to each.35 min
  104. Softmax: scores into probabilities FreeExponentiate and share out, shifting and temperature, the safe softmax, and a classifier's last step.43 min
  105. Surprise, entropy and cross-entropy FreeSurprise in bits, entropy, cross-entropy and KL divergence, perplexity, and a language model's loss.45 min
  106. KL divergence: how far one distribution is from another FreeKL by hand, cross-entropy as entropy plus KL, why it isn't symmetric, and a soft-target loss.35 min
  107. Review: find the slip (statistics and information) FreeStatistics slips to fix, mixed practice from the module, and its formula sheet.38 min
  108. Statistics and information: n, z, logits and bits FreeSample sizes from error bars, inputs from probabilities, logit gaps from softmax, and chances from bits and losses.40 min
  109. One neuron, start to finish FreeWeighted sum, activation, loss, slopes and update, by hand and in code, and what one straight line can and can't split.45 min
  110. A layer and a small network as matrices FreeTwo layers on a batch, the shape of every matrix, adding a hidden unit, and XOR solved.45 min
  111. Training a small network FreeThe backward pass in matrix form, gradient checks, a full training loop, and reading loss curves.45 min
  112. The same network in PyTorch Freenn.Sequential with the NumPy weights loaded, the same loss and slopes, then SGD and mini-batches.45 min
  113. Attention as matrix math FreeScaled dot-product scores, the causal mask, a row-by-row softmax, and blending value vectors.45 min
  114. Residual connections and layer norm FreeAdding the input back and why its slope keeps learning alive, layer norm by hand, and add and norm.40 min
  115. A transformer block, start to finish FreeA sentence through embeddings, attention, add and norm, a small network and next-word scores, every shape named.45 min
  116. Planning a model: rearranging the formulas of a working network FreeLayer widths from parameter counts, sizes from memory and compute, the attention scale, and a training plan.40 min
  117. Final test: transpose any formula FreeSlips to fix from every topic, mixed practice of every kind of rearranging, the strategy chart and the full sheet.45 min
  118. Capstone: a sign reader from the formulas FreeBuild, train and test a trail-sign classifier with NumPy alone, and report its accuracy with an error bar.45 min
Part of a free course 118 lessons

The Foundation Math For Ai & ML: from arithmetic to attention

Learn the math behind AI from arithmetic up: numbers, algebra, vectors and matrices, calculus, probability and statistics, each shown as numbers moving and checked in Python, ending with a network you build and train from the formulas.

Never trained an AI? Teach a small one to write, right in your browser. It takes about 20 seconds and needs no account. Teach it to write