`RollingMax`

Description

The RollingMax class computes the maximum value within a moving window of specified size over a sequence of data.

Initial values: The constructor requires a positive integer window_size parameter to define the rolling window.
NaN handling: NaN values are not handled natively and should be preprocessed if necessary.

Usage Example and Plot

Below is an example of using RollingMax to calculate the rolling maximum for a random dataset, along with a plot illustrating its output.

import numpy as np
import plotly.graph_objects as go
from screamer import RollingMax

# Generate example data
N = 300
window_size = 30
data = np.cumsum(np.random.normal(size=300))

# Apply rolling maximum calculation
rolling_max = RollingMax(window_size)
results = rolling_max(data)

# Plotting with Plotly
fig = go.Figure()
fig.add_trace(go.Scatter(y=data, mode='lines', name='Input Data'))
fig.add_trace(go.Scatter(y=results, mode='lines', name='Rolling Max', line=dict(color='red')))
fig.update_layout(
    title=f"Rolling Maximum with Window Size = {window_size}",
    xaxis_title="Index",
    yaxis_title="Value",
    margin=dict(l=20, r=20, t=80, b=20),
    legend=dict(orientation="h", yanchor="bottom", y=1.02, xanchor="right", x=1)
)
fig.show()

Implementation Details

ALgorithm

RollingMax used the ascending minima algorithm using a deque-based that ensures that each new maximum is calculated in O(1) constant time while using O(window_size) memory.

Complexity:

Time complexity: O(1)
Space complexity: O(window_size)

Performance

Short streams (n=1.000): 300% faster than Pandas Rolling max
Longer streams (n=1.000.000): 90% faster than Pandas Rolling max