Python 3 ImportError: cannot import name 'maxentropy'

scipy.maxentropy was removed from scipy in version 0.11 – https://docs.scipy.org/doc/scipy-0.10.1/reference/maxentropy.html

As per SciPy docs, maxentropy is removed since version 0.11.

This is what I found here:

The scipy.maxentropy module, which was deprecated in the 0.10.0 release, has been removed. Logistic regression in scikits.learn is a good and modern alternative for this functionality.

Hope it helps

Like @Gregory pointed out maxentropy was removed from scipy and they suggested to use logistic regression instead, so here is an example of max entropy spectrum estimation on a signal made of 2 sines and a random noise:

import numpy as np
from scipy import signal
from sklearn.linear_model import LogisticRegression
import matplotlib.pyplot as plt

# Generate a random time series data
np.random.seed(1234)
n = 1000
t = np.arange(n)
x = np.sin(2*np.pi*t/100) + 0.5*np.sin(2*np.pi*t/10) + 0.2*np.random.randn(n)

# Compute the autocorrelation function of the data
acf = signal.correlate(x, x, mode='full', method='auto')
acf = acf[len(acf)//2:]

# Define the constraints for maximum entropy estimation
order = 50   # order of the maximum entropy model
lags = np.arange(order)   # lags for autocorrelation function
acf_constraints = np.where(acf[:order] > 0, 1, 0)   # autocorrelation constraints
power_constraints = np.zeros(order)   # power spectrum constraints

# Fit a logistic regression model to the autocorrelation constraints
lr = LogisticRegression(penalty='none', solver='sag', max_iter=10000)
lr.fit(np.reshape(lags, (-1,1)), acf_constraints)

# Estimate the power spectrum using the logistic regression coefficients
spectrum = np.abs(np.fft.fft(lr.coef_[0]*np.hamming(order), n))**2
freq = np.fft.fftfreq(n, d=1)

# Plot the original signal and the estimated power spectrum
plt.figure(figsize=(8,6))
#plt.plot(t, x, label='Signal')
plt.plot(freq[:n//2], spectrum[:n//2], label='Power Spectrum')
plt.xlabel('Frequency')
plt.ylabel('Power')
plt.legend()
plt.show()