Index Ai Academy
Foundation · Statistics and information

Logistic regression: a line that answers yes or no

The sigmoid, the log loss, its slope p − y, a training loop, and choosing a threshold.

About 45 minutes · Free lesson

What it covers6

Why it matters

Linear regression predicts an amount. Many questions want a yes or no instead: is this email spam, is this sign an arrow, will this hiker need a rescue. A model for those should answer with a chance between 0 and 1, and a straight line can't: it runs off below 0 and above 1. Logistic regression keeps the line as a score and squeezes the score through the sigmoid, an S-shaped curve that turns any number into a chance. It is trained by maximum likelihood (lesson 100), and its loss, the log loss, has a slope so simple, p − y, that the same expression returns in nearly every classifier and language model. A single neuron with a sigmoid is exactly this model, so this lesson is also the first complete neuron of the section: score, squeeze, loss, slope, step.

What you'll be able to do

  • Turn a line's score into a chance with the sigmoid, σ(z) = 1/(1 + e⁻ᶻ), and read it off the S-curve
  • Measure a yes-or-no prediction with the log loss, and see why it punishes confident mistakes
  • Derive and use the loss's slope with respect to the score, p − y, in a short training loop
  • Train a two-feature sign detector, read it as one neuron, choose a threshold, and get the same loss and slopes from PyTorch's nn.BCEWithLogitsLoss
Quick check
Have a guess before you start. A model should give the chance a sign is an arrow. Why not use a straight line, ŷ = wx + b, as the chance?
  1. Lines are too slow to compute
  2. Chances must be whole numbers
  3. A line runs below 0 and above 1
  4. A line can't have a slope
Show the answer

A line runs below 0 and above 1

Far enough along x, any sloped line passes 1 and, the other way, drops below 0. A chance must stay between them.

The math: score, squeeze, loss

Purpose. Predict the chance of a yes, and measure how good that prediction was.

The score. As in linear regression, each example gets a score z = w·x + b: a weighted sum of its features plus a bias. A high score means "probably yes", a low one "probably no", but it can be any number.

The squeeze. The sigmoid turns the score into a chance:

z−4−2024
σ(z)0.0180.1190.50.8810.982

A score of 0 gives 0.5, a coin toss. Big positive scores approach 1 and big negative ones approach 0, never reaching either. The decision boundary is where z = 0: on one side the model leans yes, on the other no.

Quick check
From lesson 100: a model gave the right answer probability p. What is its negative log-likelihood for that one answer?
  1. 1 − p
  2. log p
  3. −log p
  4. p²
Show the answer

−log p

The NLL of one observation is −log of the probability given to what happened. The log loss below is exactly this.

The loss. If the true answer is yes (y = 1), the loss is −log p; if it's no (y = 0), −log(1 − p). So a prediction of 0.9 for a real yes costs 0.105, a shrug of 0.5 costs 0.693, and a confident 0.01 costs 4.605: the loss punishes confident mistakes far more than timid ones.

The picture. Step through one example, from features to slope. The drawing is the sigmoid's S-curve, the chance standing over every score. Watch the score land on the floor, climb to the curve to become a chance, meet the right answer (the ring), and last the gap between them: that gap is the slope.

Try it now

Then set right answer to 0 and press Use these numbers: the same chance 0.6225 now costs 0.9741, the ring drops to the floor, and the slope turns positive.

Symbols and formula.

σ(z)=11+e−zL=−[ylog⁡p+(1−y)log⁡(1−p)]∂L∂z=p−y\sigma(z) = \frac{1}{1 + e^{-z}} \qquad L = -\big[y \log p + (1 - y) \log(1 - p)\big] \qquad \frac{\partial L}{\partial z} = p - y

In plain words: squeeze the score into a chance; charge −log of the chance given to what really happened; the push on the score is just "predicted minus true".

Where p − y comes from. The chain rule of lesson 74 multiplies two links. The loss's slope with respect to p is −y/p + (1 − y)/(1 − p), and the sigmoid's slope is p(1 − p). Multiplied, almost everything cancels: −y(1 − p) + (1 − y)p = p − y. When the prediction matches the truth the push is 0; the further off, the harder the push. The slope for each weight is then (p − y)·x, lesson 81's "multiply along the path".

Quick check
What is σ(0)?
  1. 0
  2. e
  3. 1
  4. 0.5
Show the answer

0.5

1 ÷ (1 + e⁰) = 1 ÷ 2 = 0.5: a score of 0 sits exactly on the decision boundary.

In your own words
In your own words: why does the log loss punish a confident wrong answer so much more than an unsure one?

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

In Python: sigmoid, loss, slope, loop

Purpose. Write the three pieces, then train the score's w and b by gradient descent.

import numpy as np

def sigmoid(z):
    return 1 / (1 + np.exp(-z))

x = np.array([1.0, 2.0, 3.0, 4.0, 5.0, 6.0, 7.0, 8.0])
y = np.array([0.0, 0.0, 0.0, 1.0, 0.0, 1.0, 1.0, 1.0])
w, b = 0.0, 0.0
for step in range(5000):
    p = sigmoid(w * x + b)
    w = w - 0.1 * np.mean((p - y) * x)
    b = b - 0.1 * np.mean(p - y)
p = sigmoid(w * x + b)
print(round(w, 3), round(b, 3), round(-b / w, 2))
print(p.round(2).tolist())
It runs right here in your browser, with nothing to install.

Eight readings, x from 1 to 8, with the answer mostly no below 4.5 and mostly yes above. Training settles at w = 1.271 and b = −5.714, so the boundary, where z = 0, sits at x = 5.714 ÷ 1.271 = 4.5. Two points are on the "wrong" side, x = 4 (a yes) and x = 5 (a no): no single boundary can separate them, so the model gives them middling chances, 0.35 and 0.65, rather than confident wrong ones.

Code flow. Here is the same loop on four points, three passes, with every piece on its own line: the score, the squeeze, the loss, the slope p − y, and the two updates. Step through and watch the boxes: p is one chance per point, and dz is how far each one is from its answer.

A named mistake: logging 0. If a prediction reaches exactly 0 or 1, np.log(0) returns minus infinity and the loss becomes infinite or NaN. Floats get there sooner than you'd think: σ(40) rounds to exactly 1.0, so 1 − p is 0.

import numpy as np

p = 1 / (1 + np.exp(-40.0))
print(p == 1.0, 1 - p)

Libraries guard against it by clipping p into [ε, 1 − ε] before the log, or by computing the loss straight from the score in a stable form, as PyTorch does in the AI part. When a training loss suddenly prints nan, a log of 0 is the first suspect.

Worked example

Worked example: the loss for good and bad predictions

Compute the log loss for five predictions of a real yes, from confident and right to confident and wrong.

import numpy as np

p = np.array([0.99, 0.9, 0.5, 0.1, 0.01])
loss = -np.log(p)
print(loss.round(3).tolist())
print(round(float(loss.mean()), 3))
In the lesson you can change this and run it.

The losses run 0.010, 0.105, 0.693, 2.303 and 4.605. Going from 0.99 to 0.9 costs a tenth of a point; going from 0.1 to 0.01 costs 2.3 points. Averaged, the one confident mistake supplies more than half the total. That lopsidedness is deliberate: it's what drives a classifier away from confident errors first, and it's why a model that says 0.5 for everything still beats one that is sometimes sure and wrong.

Quick check
What does this print?
import numpy as np

z = np.array([-4.0, -2.0, 0.0, 2.0, 4.0])
print((1 / (1 + np.exp(-z))).round(3).tolist())
  1. [-4.0, -2.0, 0.0, 2.0, 4.0]
  2. [0.018, 0.119, 0.5, 0.881, 0.982]
  3. [0.0, 0.0, 0.5, 1.0, 1.0]
  4. [0.982, 0.881, 0.5, 0.119, 0.018]
Show the answer

[0.018, 0.119, 0.5, 0.881, 0.982]

The math part's table: below a half for negative scores, a half at 0, above a half for positive ones, and σ(−z) = 1 − σ(z).

Draw it. The trained model from the first cell, w = 1.271 and b = −5.714, as a curve over x, with the eight readings as dots. The third argument, "o", tells plt.plot to draw dots instead of a line:

import numpy as np
import matplotlib.pyplot as plt

xs = np.linspace(0, 9, 91)
plt.plot(xs, 1 / (1 + np.exp(-(1.271 * xs - 5.714))))
plt.plot([1, 2, 3, 4, 5, 6, 7, 8], [0, 0, 0, 1, 0, 1, 1, 1], "o")
plt.xlabel("x")
plt.ylabel("chance of yes")
plt.show()

The math part's S-curve again, stretched and slid along x by the weight and the bias.

Quick check
Read the plot. Where does the curve cross a chance of 0.5?
  1. At x = 4.5, between the two mixed points
  2. At x = 5.714, the bias
  3. At x = 0, where the score is 0
  4. At x = 1.271, the weight
Show the answer

At x = 4.5, between the two mixed points

The chance is 0.5 where the score 1.271x − 5.714 is 0: x = 5.714 ÷ 1.271 = 4.5, the boundary the first cell printed. It sits between x = 4 (a yes) and x = 5 (a no).

Drill

Train the boundary

Fill in the two update lines of this training loop with the slopes (p − y)·x and (p − y), averaged, and learning rate 0.1. Then set boundary to −b/w, rounded to 2 places, and acc to the share of points the model gets right at threshold 0.5.

import numpy as np

def sigmoid(z):
    return 1 / (1 + np.exp(-z))

x = np.array([1.0, 2.0, 3.0, 4.0, 5.0, 6.0, 7.0, 8.0])
y = np.array([0.0, 0.0, 0.0, 1.0, 0.0, 1.0, 1.0, 1.0])
w, b = 0.0, 0.0
for step in range(5000):
    p = sigmoid(w * x + b)
    w = w
    b = b
boundary = 0
acc = 0
print(boundary, acc)
In the lesson, your code is checked as you write it.
In your own words
The trained model gives x = 4 and x = 5 chances of 0.35 and 0.65 instead of confident answers. Why is that the right behavior?

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

In AI and ML: a neuron, and a threshold

Purpose. See logistic regression as one neuron, and choose where to turn its chance into a decision.

A neuron. A neuron with a sigmoid computes a weighted sum of its inputs plus a bias, then squeezes it: exactly z = w·x + b and p = σ(z). Trained with the log loss, it is logistic regression. A network stacks many of them, and its last layer for a yes-or-no question is still this neuron. The slope p − y is where backpropagation (lesson 81) starts.

Is this sign an arrow? The trail club's sign camera describes each sign by two numbers from 0 to 1: how pointed its tip is and how wide it is. Arrows are pointed and narrow; tent-site boards are blunt and wide. Ten signs, five of each, trained like the loop above, now with two weights, so the score is one dot product per sign, X @ w + b, and each weight's slope is X.T @ (p - y) averaged:

import numpy as np

X = np.array([[0.8, 0.1], [0.0, 0.7], [0.5, 0.3], [0.5, 0.6], [0.9, 0.1],
              [0.6, 0.6], [0.4, 0.3], [0.1, 0.9], [0.9, 0.2], [0.0, 0.6]])
y = np.array([1, 0, 0, 1, 1, 0, 1, 0, 1, 0])
w, b = np.zeros(2), 0.0
for step in range(2000):
    p = 1 / (1 + np.exp(-(X @ w + b)))
    w = w - 1.0 * X.T @ (p - y) / len(y)
    b = b - 1.0 * np.mean(p - y)
p = 1 / (1 + np.exp(-(X @ w + b)))
loss = np.mean(-(y * np.log(p) + (1 - y) * np.log(1 - p)))
print(w.round(2).tolist(), round(b, 2), round(loss, 3))
print(p.round(2).tolist())

Pointed counts for an arrow (w₁ = 4.77) and wide counts against (w₂ = −4.42), with a bias of −0.37 and a mean loss of 0.359. The chances run from 0.97 for a sharp narrow arrow down to 0.02 for a wide blunt board. Two signs are hard: sign 3, a narrow board with a slanted end, gets 0.67, and sign 4, a wide arrow, only 0.35.

Choosing a threshold. To act, a threshold turns each chance into a yes or a no. At 0.5, sign 3 (0.67) is called an arrow though it's a board, and sign 4 (0.35) is called a board though it's an arrow: 8 of 10 right, with one mistake of each kind. Move the threshold and the count can stay the same while the mistakes change places. The threshold doesn't change what the model knows; it chooses which mistake is cheaper: a missed arrow, or a false one. For a rescue alert you'd lower it; for a spam filter that deletes mail, you'd raise it. Lesson 103 names the two kinds of mistake and scores each.

Quick check
What does this print?
import numpy as np

p = np.array([0.9, 0.6, 0.4, 0.2])
print((p >= 0.5).astype(int).tolist(), (p >= 0.3).astype(int).tolist())
  1. [1, 1, 1, 0] [1, 1, 0, 0]
  2. [1, 1, 0, 0] [1, 1, 1, 0]
  3. [1, 1, 0, 0] [1, 1, 0, 0]
  4. [0, 0, 1, 1] [0, 1, 1, 1]
Show the answer

[1, 1, 0, 0] [1, 1, 1, 0]

Lowering the threshold from 0.5 to 0.3 turns the 0.4 into a yes: more yes calls, the same chances.

Quick check
Raising a classifier's threshold from 0.5 to 0.9 usually does what?
  1. Nothing, since the chances stay the same
  2. Fewer yes calls: fewer false alarms, more misses
  3. More yes calls
  4. Changes the model's weights
Show the answer

Fewer yes calls: fewer false alarms, more misses

Only the surest cases pass 0.9, so it says yes less often: its yes calls are more often right (precision, from lesson 90), but it finds fewer of the real cases (recall).

In PyTorch. PyTorch's loss for this model is nn.BCEWithLogitsLoss: binary cross-entropy, the log loss, taken straight from the score (the logit), with the squeeze done inside in a form that never logs a 0. The cell loads the weights the NumPy cell printed. requires_grad_() asks PyTorch to track the scores, and backward() finds the loss's slope for each sign's score. The loss is a mean over 10 signs, so each slope arrives divided by 10:

import torch
import torch.nn as nn

X = torch.tensor([[0.8, 0.1], [0.0, 0.7], [0.5, 0.3], [0.5, 0.6], [0.9, 0.1],
                  [0.6, 0.6], [0.4, 0.3], [0.1, 0.9], [0.9, 0.2], [0.0, 0.6]])
y = torch.tensor([1.0, 0.0, 0.0, 1.0, 1.0, 0.0, 1.0, 0.0, 1.0, 0.0])
z = X @ torch.tensor([4.77, -4.42]) - 0.37
z.requires_grad_()
loss = nn.BCEWithLogitsLoss()(z, y)
loss.backward()
print(round(loss.item(), 3))
slope = z.grad * len(y)
print(torch.allclose(slope, torch.sigmoid(z) - y))
print(round(slope[2].item(), 2), round(slope[3].item(), 2))

The loss is 0.359, the mean loss the NumPy cell printed. True says every slope is exactly p − y, the formula from the math part. The biggest pushes go to the two hard signs: +0.67 pushes board 3's score down, −0.65 pushes arrow 4's up.

Drill

Mistakes at three thresholds

The detector's ten chances are in p and the real answers in truth. For thresholds 0.3, 0.5 and 0.7, in that order, set accs to the accuracy, misses to the number of real arrows called no, and false_alarms to the number of boards called yes, each a list of three.

import numpy as np

p = np.array([0.95, 0.03, 0.67, 0.35, 0.97, 0.46, 0.55, 0.02, 0.95, 0.05])
truth = np.array([1, 0, 0, 1, 1, 0, 1, 0, 1, 0])
accs = []
misses = []
false_alarms = []
print(accs, misses, false_alarms)
In the lesson, your code is checked as you write it.
In your own words
For a model that flags hikers who may need rescue, would you set the threshold above or below 0.5? Why?

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 does the sigmoid do in logistic regression?

  1. Turns a score into a chance
  2. Picks the features
  3. Sets the threshold for a yes
  4. Measures the loss

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