Backpropagation

When the output is wrong, how much is each weight to blame?

A tiny network, one training example, and every number shown. The table at the bottom is the heart of how all neural networks learn: for each weight, backprop multiplies a few local slopes together to get "how much does the error change if I nudge this weight?"

The network after 2335 updates

w1 = 0.319 w2 = 0.537 w3 = 0.417 w4 = 0.635 w5 = −3.408 w6 = −3.380 w7 = 2.380 w8 = 2.441 i1 0.050 i2 0.100 h1 0.603 h2 0.607 o1 0.029 o2 0.971 target 0.01 target 0.99 bias b1 = 0.35 output bias b2 = 0.6

Train it

0.5
0.000359total error E
0.0291, 0.9712outputs (want 0.01, 0.99)
0.00000.0755840.1491,1680.2241,7510.2982,335updates

1. Forward pass: compute the answer

net_h1 = w1·i1 + w2·i2 + b1 = 0.3186·0.05 + 0.5373·0.10 + 0.35 = 0.419657
h1 = σ(net_h1) = 0.603401    h2 = σ(0.434316) = 0.606904
net_o1 = w5·h1 + w6·h2 + b2 = −3.507789    o1 = 0.029091
net_o2 = w7·h1 + w8·h2 + b2 = 3.517918    o2 = 0.971193
E = ½(t1 − o1)² + ½(t2 − o2)² = 0.000182 + 0.000177 = 0.000359087

What to notice

Output weights are easy. w5's blame is just three numbers multiplied: how wrong o1 is, how steep o1's sigmoid is, and how big h1 was (a big input means the weight mattered more).

Hidden weights reuse that work. w1 can't see the error directly. Its blame arrives through both outputs: output 1's blame times w5, plus output 2's blame times w7. Then the same local slopes. That reuse is what makes backprop fast enough for billions of weights.

The check column agrees. "Nudge and measure" gives the same numbers without any calculus, but it needs two extra forward passes per weight. For an LLM that would be billions of passes per step; backprop needs one.

2. Backward pass: blame for every weight

WeightLocal slopes multiplied together (chain rule)∂E/∂wCheck (nudge)New value = w − 0.50 × ∂E/∂w
w5 (o₁ − t₁) 0.01909 × o₁(1 − o₁) 0.02825 × h₁ 0.60340 0.0003254 0.0003254 −3.40774 → −3.40789826
w6 (o₁ − t₁) 0.01909 × o₁(1 − o₁) 0.02825 × h₂ 0.60690 0.0003273 0.0003273 −3.38037 → −3.38053086
w7 (o₂ − t₂) −0.01881 × o₂(1 − o₂) 0.02798 × h₁ 0.60340 −0.0003175 −0.0003175 2.38026 → 2.38041923
w8 (o₂ − t₂) −0.01881 × o₂(1 − o₂) 0.02798 × h₂ 0.60690 −0.0003193 −0.0003193 2.44135 → 2.44151175
w1 δo₁·w5 + δo₂·w7 −0.00309 × h₁(1 − h₁) 0.23931 × i₁ 0.05000 −0.0000370 −0.0000370 0.31863 → 0.31864757
w2 δo₁·w5 + δo₂·w7 −0.00309 × h₁(1 − h₁) 0.23931 × i₂ 0.10000 −0.0000739 −0.0000739 0.53726 → 0.53729514
w3 δo₁·w6 + δo₂·w8 −0.00311 × h₂(1 − h₂) 0.23857 × i₁ 0.05000 −0.0000371 −0.0000371 0.41726 → 0.41728183
w4 δo₁·w6 + δo₂·w8 −0.00311 × h₂(1 − h₂) 0.23857 × i₂ 0.10000 −0.0000741 −0.0000741 0.63453 → 0.63456366

At the start with learning rate 0.5 you should see w5 → 0.35891648 and w1 → 0.14978072, the same as the original worked example. δo means (o − t)·o(1 − o), the output neuron's blame.

The code that just ran

Everything on this page came from calling this code, which lives at the top of this same file.

Show or hide the source
final class TinyNet
{
    public array $w = ['w1' => 0.15, 'w2' => 0.20, 'w3' => 0.25, 'w4' => 0.30,
                       'w5' => 0.40, 'w6' => 0.45, 'w7' => 0.50, 'w8' => 0.55,
                       'b1' => 0.35, 'b2' => 0.60];   // biases stay fixed, as in the original example
    public array $in = [0.05, 0.10];
    public array $target = [0.01, 0.99];

    public function forward(): array
    {
        $w = $this->w; [$i1, $i2] = $this->in; [$t1, $t2] = $this->target;
        $nh1 = $w['w1'] * $i1 + $w['w2'] * $i2 + $w['b1'];  $h1 = sigmoid($nh1);
        $nh2 = $w['w3'] * $i1 + $w['w4'] * $i2 + $w['b1'];  $h2 = sigmoid($nh2);
        $no1 = $w['w5'] * $h1 + $w['w6'] * $h2 + $w['b2'];  $o1 = sigmoid($no1);
        $no2 = $w['w7'] * $h1 + $w['w8'] * $h2 + $w['b2'];  $o2 = sigmoid($no2);
        $E1 = 0.5 * ($t1 - $o1) ** 2;
        $E2 = 0.5 * ($t2 - $o2) ** 2;
        return compact('i1', 'i2', 't1', 't2', 'nh1', 'h1', 'nh2', 'h2', 'no1', 'o1', 'no2', 'o2', 'E1', 'E2') + ['E' => $E1 + $E2];
    }

    public function loss(): float { return $this->forward()['E']; }

    // BACKPROPAGATION. Each gradient is a product of simple local slopes:
    //   (o − t)      how the error changes with the output
    //   o·(1 − o)    how the sigmoid output changes with its input
    //   h or i       how that input changes with the weight
    // Hidden weights get the SUM of the blame flowing back from both outputs.
    public function gradients(): array
    {
        $f = $this->forward(); $w = $this->w;
        $do1 = ($f['o1'] - $f['t1']) * $f['o1'] * (1 - $f['o1']);   // δ of output 1
        $do2 = ($f['o2'] - $f['t2']) * $f['o2'] * (1 - $f['o2']);   // δ of output 2
        $back1 = $do1 * $w['w5'] + $do2 * $w['w7'];                 // blame reaching h1
        $back2 = $do1 * $w['w6'] + $do2 * $w['w8'];                 // blame reaching h2
        $s1 = $f['h1'] * (1 - $f['h1']);
        $s2 = $f['h2'] * (1 - $f['h2']);

        $g = fn($factors) => ['factors' => $factors, 'value' => array_product(array_column($factors, 1))];
        return [
            'w5' => $g([['(o₁ − t₁)', $f['o1'] - $f['t1']], ['o₁(1 − o₁)', $f['o1'] * (1 - $f['o1'])], ['h₁', $f['h1']]]),
            'w6' => $g([['(o₁ − t₁)', $f['o1'] - $f['t1']], ['o₁(1 − o₁)', $f['o1'] * (1 - $f['o1'])], ['h₂', $f['h2']]]),
            'w7' => $g([['(o₂ − t₂)', $f['o2'] - $f['t2']], ['o₂(1 − o₂)', $f['o2'] * (1 - $f['o2'])], ['h₁', $f['h1']]]),
            'w8' => $g([['(o₂ − t₂)', $f['o2'] - $f['t2']], ['o₂(1 − o₂)', $f['o2'] * (1 - $f['o2'])], ['h₂', $f['h2']]]),
            'w1' => $g([['δo₁·w5 + δo₂·w7', $back1], ['h₁(1 − h₁)', $s1], ['i₁', $f['i1']]]),
            'w2' => $g([['δo₁·w5 + δo₂·w7', $back1], ['h₁(1 − h₁)', $s1], ['i₂', $f['i2']]]),
            'w3' => $g([['δo₁·w6 + δo₂·w8', $back2], ['h₂(1 − h₂)', $s2], ['i₁', $f['i1']]]),
            'w4' => $g([['δo₁·w6 + δo₂·w8', $back2], ['h₂(1 − h₂)', $s2], ['i₂', $f['i2']]]),
        ];
    }

    // Sanity check without any calculus: nudge the weight up and down a hair
    // and see how much the loss moves. Should match backprop almost exactly.
    public function numericalGradient(string $name, float $eps = 1e-5): float
    {
        $keep = $this->w[$name];
        $this->w[$name] = $keep + $eps; $up = $this->loss();
        $this->w[$name] = $keep - $eps; $down = $this->loss();
        $this->w[$name] = $keep;
        return ($up - $down) / (2 * $eps);
    }

    public function step(float $lr): void
    {
        foreach ($this->gradients() as $name => $g) $this->w[$name] -= $lr * $g['value'];
    }
}