diff --git a/CHANGELOG b/CHANGELOG
index 51a4969..b01c9d4 100644
--- a/CHANGELOG
+++ b/CHANGELOG
@@ -2,7 +2,8 @@
 All notable changes to this project will be documented in this file.
 
 
-## [Unreleased]
+## [0.4] - 2023-09-01
+# Blog post: TBA
 ### Added
 - nested: NS-SMC sampler of Salomone et al (2018)
 - datasets: Liver
diff --git a/README.md b/README.md
index 84c06de..4ae492d 100644
--- a/README.md
+++ b/README.md
@@ -1,9 +1,9 @@
+![logo](logo.png)
+
 # particles #
 
 Sequential Monte Carlo in python. 
 
-![logo](logo.png)
-
 ## Motivation ##
 
 This package was developed to complement the following book:
diff --git a/papers/complexity_smoothing/mle_neuro_nrsteps.py b/papers/complexity_smoothing/mle_neuro_nrsteps.py
index 2bd3bda..ceaec81 100644
--- a/papers/complexity_smoothing/mle_neuro_nrsteps.py
+++ b/papers/complexity_smoothing/mle_neuro_nrsteps.py
@@ -9,19 +9,12 @@
 """
 
 
-import itertools
 import time
-import pickle
 
-import matplotlib
-import numpy as np
 from matplotlib import pyplot as plt
 import seaborn as sb
 
 import particles
-from particles import datasets as dts
-from particles import distributions as dists
-from particles import state_space_models as ssms
 
 import mle_neuro
 
@@ -29,21 +22,21 @@ def worker(rho0=0.1, sig20=0.5, N=10, ffbs='mcmc', maxiter=1, mcmc_steps=1):
     tic = time.perf_counter()
     results = mle_neuro.EM(rho0, sig20, N=N, maxiter=maxiter, xatol=0., 
                            ffbs=ffbs, mcmc_steps=mcmc_steps)
+    results['cpu_time'] = time.perf_counter() - tic
     results['ffbs'] = ffbs
     results['mcmc_steps'] = mcmc_steps
-    cpu_time = time.perf_counter() - tic
-    results['cpu_time'] = cpu_time
     return results
 
 nruns = 100
 maxiter = 20
-N = 1000  # TODO was 100
+N = 1000
 nsteps = [1, 3]
 
 results = particles.utils.multiplexer(f=worker, N=N, maxiter=maxiter,
-                                      mcmc_steps=nsteps, nruns=nruns, nprocs=0)
+                                      mcmc_steps=nsteps, nruns=nruns, nprocs=1)
+ref_method = 'hybrid'  # replace by 'pureject' to see improvement over book xp
 ref_results = particles.utils.multiplexer(f=worker, N=N, maxiter=maxiter,
-                                          ffbs='hybrid', nruns=nruns, nprocs=0)
+                                          ffbs=ref_method, nruns=nruns, nprocs=1)
 results.extend(ref_results)
 
 flat_res = []