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.
""" anyplot.ai
line-loss-training: Training Loss Curve
Library: letsplot 4.9.0 | Python 3.13.13
Quality: 92/100 | Updated: 2026-05-14
"""
import os
import shutil
from pathlib import Path
import numpy as np
import pandas as pd
from lets_plot import (
LetsPlot,
aes,
element_blank,
element_line,
element_rect,
element_text,
geom_line,
geom_point,
geom_vline,
ggplot,
ggsize,
labs,
scale_color_manual,
theme,
theme_minimal,
)
from lets_plot.export import ggsave
LetsPlot.setup_html()
# Theme tokens
THEME = os.getenv("ANYPLOT_THEME", "light")
PAGE_BG = "#FAF8F1" if THEME == "light" else "#1A1A17"
ELEVATED_BG = "#FFFDF6" if THEME == "light" else "#242420"
INK = "#1A1A17" if THEME == "light" else "#F0EFE8"
INK_SOFT = "#4A4A44" if THEME == "light" else "#B8B7B0"
# Okabe-Ito palette
BRAND = "#009E73" # Training loss (first series)
ACCENT = "#C475FD" # Validation loss (second series)
OPTIMAL = "#DC2626" # Optimal epoch marker
# Data - Simulated neural network training loss over 100 epochs
np.random.seed(42)
epochs = np.arange(1, 101)
# Training loss: starts high, decreases with noise - continues to decrease throughout
train_loss = 2.5 * np.exp(-0.05 * epochs) + 0.08 + np.random.normal(0, 0.015, len(epochs))
# Validation loss: decreases then increases (overfitting after ~50 epochs)
val_loss_base = 2.5 * np.exp(-0.045 * epochs) + 0.2
noise = np.random.normal(0, 0.02, len(epochs))
val_loss = val_loss_base + noise
# Add overfitting effect - validation loss increases after epoch 50
overfitting_start = 50
val_loss[overfitting_start:] = val_loss[overfitting_start:] + 0.008 * (epochs[overfitting_start:] - overfitting_start)
# Find optimal epoch (minimum validation loss)
optimal_epoch = int(epochs[np.argmin(val_loss)])
optimal_loss = float(val_loss.min())
# Create DataFrame for plotting
df = pd.DataFrame(
{
"Epoch": np.tile(epochs, 2),
"Loss": np.concatenate([train_loss, val_loss]),
"Type": ["Training Loss"] * len(epochs) + ["Validation Loss"] * len(epochs),
}
)
# Optimal point marker
optimal_df = pd.DataFrame({"Epoch": [optimal_epoch], "Loss": [optimal_loss]})
# Custom theme
custom_theme = theme(
plot_background=element_rect(fill=PAGE_BG, color=PAGE_BG),
panel_background=element_rect(fill=PAGE_BG, color=PAGE_BG),
panel_border=element_blank(),
panel_grid_major=element_line(color=INK_SOFT, size=0.3, linetype="solid"),
panel_grid_minor=element_blank(),
axis_line_x=element_line(color=INK_SOFT, size=0.5),
axis_line_y=element_line(color=INK_SOFT, size=0.5),
axis_title=element_text(size=20, color=INK),
axis_text=element_text(size=16, color=INK_SOFT),
plot_title=element_text(size=24, color=INK),
legend_background=element_rect(fill=ELEVATED_BG, color=INK_SOFT),
legend_text=element_text(size=16, color=INK_SOFT),
legend_title=element_text(size=16, color=INK),
legend_position="top",
)
# Create plot
plot = (
ggplot(df, aes(x="Epoch", y="Loss", color="Type"))
+ geom_line(size=1.5, alpha=0.9)
+ geom_point(data=optimal_df, mapping=aes(x="Epoch", y="Loss"), color=OPTIMAL, size=6, shape=18, inherit_aes=False)
+ geom_vline(xintercept=optimal_epoch, color=OPTIMAL, size=0.8, linetype="dashed", alpha=0.7)
+ scale_color_manual(values=[BRAND, ACCENT])
+ labs(title="line-loss-training · letsplot · anyplot.ai", x="Epoch", y="Cross-Entropy Loss", color="")
+ theme_minimal()
+ custom_theme
+ ggsize(1600, 900)
)
# Save as PNG (scale 3x for 4800x2700)
ggsave(plot, f"plot-{THEME}.png", scale=3)
ggsave(plot, f"plot-{THEME}.html")
# Move files from lets-plot-images directory to current directory
images_dir = Path("lets-plot-images")
if images_dir.exists():
for file in images_dir.glob(f"plot-{THEME}.*"):
shutil.move(str(file), str(file.name))
Part of Training Loss Curve on anyplot.ai.