Training Loss Curve — Matplotlib

A line plot showing training and validation loss curves over epochs during neural network training. This visualization is essential for monitoring model training, detecting overfitting (when validation loss diverges from training loss), and determining optimal early stopping points. The dual-curve display reveals the gap between training and generalization performance.

Training Loss Curve rendered with Matplotlib

Python source (Matplotlib)

""" anyplot.ai
line-loss-training: Training Loss Curve
Library: matplotlib 3.10.9 | Python 3.13.13
Quality: 90/100 | Updated: 2026-05-14
"""

import os

import matplotlib.pyplot as plt
import numpy as np


# Theme tokens
THEME = os.getenv("ANYPLOT_THEME", "light")
PAGE_BG = "#FAF8F1" if THEME == "light" else "#1A1A17"
INK = "#1A1A17" if THEME == "light" else "#F0EFE8"
INK_SOFT = "#4A4A44" if THEME == "light" else "#B8B7B0"

# Okabe-Ito colors
TRAINING_COLOR = "#009E73"
VALIDATION_COLOR = "#C475FD"

# Data
np.random.seed(42)
epochs = np.arange(1, 151)

# Training loss: sigmoid saturation pattern (different from exponential)
# Starts high (~2.2) and rapidly decays, then plateaus
train_loss = 2.2 / (1 + np.exp((epochs - 25) / 8)) + 0.12 + np.random.randn(150) * 0.012

# Validation loss: similar initial decay but with step-function overfitting
# Decays to minimum around epoch 65, then increases due to overfitting
val_base = 2.1 / (1 + np.exp((epochs - 30) / 9)) + 0.18
val_loss = np.copy(val_base) + np.random.randn(150) * 0.015

# Add step-function overfitting: sharp increase after epoch 65
overfitting_start = 65
val_loss[overfitting_start:] += (
    np.linspace(0, 0.5, 150 - overfitting_start) + np.random.randn(150 - overfitting_start) * 0.012
)

# Find minimum validation loss epoch
min_val_epoch = np.argmin(val_loss) + 1
min_val_loss = val_loss[min_val_epoch - 1]

# Plot
fig, ax = plt.subplots(figsize=(16, 9), facecolor=PAGE_BG)
ax.set_facecolor(PAGE_BG)

ax.plot(epochs, train_loss, linewidth=3, color=TRAINING_COLOR, label="Training Loss", marker="o", markersize=4)
ax.plot(epochs, val_loss, linewidth=3, color=VALIDATION_COLOR, label="Validation Loss", marker="s", markersize=4)

# Mark minimum validation loss (optimal early stopping)
ax.scatter([min_val_epoch], [min_val_loss], s=250, color=TRAINING_COLOR, zorder=5, edgecolors=PAGE_BG, linewidth=1.5)
ax.annotate(
    f"Epoch {min_val_epoch}",
    xy=(min_val_epoch, min_val_loss),
    xytext=(min_val_epoch + 20, min_val_loss - 0.25),
    fontsize=14,
    color=INK,
    arrowprops={"arrowstyle": "->", "color": INK_SOFT, "lw": 1.5},
    bbox={"boxstyle": "round,pad=0.4", "facecolor": PAGE_BG, "edgecolor": INK_SOFT, "linewidth": 1},
)

# Style
ax.set_xlabel("Epoch", fontsize=20, color=INK)
ax.set_ylabel("Cross-Entropy Loss", fontsize=20, color=INK)
ax.set_title("line-loss-training · matplotlib · anyplot.ai", fontsize=24, fontweight="medium", color=INK)
ax.tick_params(axis="both", labelsize=16, colors=INK_SOFT)

# Spines
ax.spines["top"].set_visible(False)
ax.spines["right"].set_visible(False)
for s in ("left", "bottom"):
    ax.spines[s].set_color(INK_SOFT)

# Grid (subtle, y-axis only)
ax.yaxis.grid(True, alpha=0.15, linewidth=0.8, color=INK)

# Legend (upper left to avoid overlap)
leg = ax.legend(fontsize=16, loc="upper left", framealpha=0.95)
if leg:
    leg.get_frame().set_facecolor(PAGE_BG)
    leg.get_frame().set_edgecolor(INK_SOFT)
    leg.get_frame().set_linewidth(0.8)
    for text in leg.get_texts():
        text.set_color(INK_SOFT)

ax.set_xlim(0, 155)
plt.tight_layout()
plt.savefig(f"plot-{THEME}.png", dpi=300, bbox_inches="tight", facecolor=PAGE_BG)

Part of Training Loss Curve on anyplot.ai.

Other implementations