test_reference_extended

A $5 \times 5$ matrix with float values. We will scan row by row to find the maximum.

Scanning row 0: values are $0.2, 0.5, 0.1, 0.8, 0.3$. Current max is $0.8$ at cell $(0,3)$.

Scanning row 1: values are $0.6, 0.4, 0.9, 0.2, 0.7$. New max is $0.9$ at cell $(1,2)$.

Scanning row 2: values are $0.1, 0.3, 0.5, 0.4, 0.6$. Max remains $0.9$ at $(1,2)$.

Scanning row 3: values are $0.7, 0.2, 0.4, 0.1, 0.8$. Max remains $0.9$ at $(1,2)$.

Scanning row 4: values are $0.3, 0.6, 0.5, 0.7, 0.4$. Max remains $0.9$ at $(1,2)$.

Scan complete. The maximum value $0.9$ is at cell $(1,2)$. Marked as good.

An empty $2$D coordinate plane with grid and axes from $-5$ to $5$.

Add point $A = (-3, 2)$.

Add point $B = (1, 4)$.

Add point $C = (2, -1)$.

Add point $D = (4, 3)$.

Add point $E = (-2, -3)$.

Add line $y = x + 1$ (slope $1$, intercept $1$). Points $B = (1,4)$ lies above; $C = (2,-1)$ lies below.

Begin tracking three metrics: loss, accuracy, and lr (learning rate).

Step 1: loss $= 2.5$, accuracy $= 0.1$, lr $= 0.01$.

Step 2: loss drops to $1.8$, accuracy rises to $0.3$.

Step 3: loss $= 1.2$, accuracy $= 0.55$, lr halved to $0.005$.

Step 4: loss $= 0.7$, accuracy $= 0.72$. Steady improvement.

Step 5: loss $= 0.4$, accuracy $= 0.85$, lr reduced to $0.001$.

Step 6: final loss $= 0.25$, accuracy $= 0.91$. Training converging nicely.