Skip to content

Commit

Permalink
start work on tidying up doenut.py
Browse files Browse the repository at this point in the history
  • Loading branch information
@cwoac
cwoac committed Nov 1, 2023
1 parent b49bec6 commit b8138da
Showing 1 changed file with 29 additions and 36 deletions.
65 changes: 29 additions & 36 deletions doenut/doenut.py
View file
Original file line number Diff line number Diff line change
@@ -1,24 +1,18 @@
#############################################################
# #
# DOENUT #
# Design Of Experiments ...
# #
#############################################################
"""
DOENUT
Design of Experiments ...
"""

# first we import some useful libraries
import numpy as np
import pandas as pd
import copy
import random
from matplotlib import pyplot as plt
from sklearn.linear_model import LinearRegression
from sklearn.pipeline import make_pipeline
from sklearn.preprocessing import PolynomialFeatures
from sklearn import preprocessing


def orthogonal_scaling(inputs):
## the scaling thingy that Modde uses
# the scaling thingy that Modde uses
inputs_max = np.max(inputs)
inputs_min = np.min(inputs)
Mj = (inputs_min + inputs_max) / 2
Expand All @@ -36,7 +30,7 @@ def scale_1D_data(scaler, data, do_fit=True):


def applying_orthogonal_scaling_to_new_data(new_data):
## the scaling thingy that Modde uses
# the scaling thingy that Modde uses
new_data = (new_data - Mj) / Rj
return new_data

Expand All @@ -53,7 +47,7 @@ def find_replicates(inputs):
def replicate_plot(inputs, responses, key):
"""Plots a replicate plot which shows all experiments
and identifies the replicates
inputs:
inputs:
responses:
key: column in responses that you wish to plot"""
plt.title(key)
Expand Down Expand Up @@ -103,7 +97,7 @@ def train_model(inputs,
verbose: this is a common setting in coding, if true it means run in verbose mode and many bits of information are printed to the screen."""
if do_scaling_here:
inputs, _, _ = orthogonal_scaling(inputs)
if test_responses == None:
if test_responses is None:
test_responses = responses
model = LinearRegression(fit_intercept=fit_intercept)
model.fit(inputs, responses)
Expand Down Expand Up @@ -139,7 +133,7 @@ def plot_observed_vs_predicted(responses,

def Calculate_R2(ground_truth, predictions, key, word='test', verbose=True):
"""Calculates R2 from input data
You can use this to calculate q2 if you're
You can use this to calculate q2 if you're
using the test ground truth as the mean
else use calculate Q2
I think this is what Modde uses for PLS fitting"""
Expand Down Expand Up @@ -171,7 +165,7 @@ def Calculate_Q2(ground_truth,
word='test',
verbose=True):
"""A different way of calculating Q2
this uses the mean from the training data, not the
this uses the mean from the training data, not the
test ground truth"""
errors = ground_truth[[key]] - predictions[[key]]
train_mean = np.mean(train_responses[[key]], axis=0)
Expand All @@ -186,7 +180,7 @@ def Calculate_Q2(ground_truth,
f'Sum of squares of the residuals (explained variance) is {sum_squares_residuals}'
)
sum_squares_total = sum((ground_truth[key] - train_mean[0])**2)
##### stuff from Modde
### stuff from Modde
#errors/1

if verbose:
Expand Down Expand Up @@ -277,7 +271,7 @@ def calc_averaged_model(inputs,
fit_intercept=False,
do_scaling_here=False,
use_scaled_inputs=False):
"""Uses 'leave one out' method to train and test a series
"""Uses 'leave one out' method to train and test a series
of models
inputs: full set of terms for the model (x_n)
responses: responses to model (ground truth, y)
Expand Down Expand Up @@ -306,8 +300,7 @@ def calc_averaged_model(inputs,
inputs, _, _ = orthogonal_scaling(inputs)
whole_inputs = inputs
whole_responses = responses
if (drop_duplicates == 'Yes') or (type(drop_duplicates) == bool
and drop_duplicates):
if (drop_duplicates == 'Yes') or (type(drop_duplicates) == bool and drop_duplicates):
print('Dropping replicates')
duplicates = [x for x in whole_inputs[whole_inputs.duplicated()].index]
elif drop_duplicates == 'average':
Expand Down Expand Up @@ -379,7 +372,7 @@ def calc_averaged_model(inputs,
Q2 = Calculate_Q2(
ground_truth=df_GT,
predictions=df_pred,
train_responses=whole_responses, ### is it this one?
train_responses=whole_responses, # is it this one?
word='test',
key=key,
verbose=True)
Expand All @@ -406,7 +399,7 @@ def calc_ave_coeffs_and_errors(coeffs, labels, errors='std', normalise=False):
# this is an approximation assuming a gaussian distribution in your coeffs
error_bars = 2 * stds
else:
printf(f'Error: errors setting {errors} not known, chose std or p95')
print(f'Error: errors setting {errors} not known, chose std or p95')

return ave_coeffs, error_bars

Expand Down Expand Up @@ -467,12 +460,12 @@ def calulate_R2_and_Q2_for_models(inputs,
print(f"Input terms are {input_column_list}")
print(f"Input Responses are {response_column_list}\n")
# make a linear regression model
if response_selector == None:
if response_selector is None:
# do all columns
res_col_num_list = [x for x in range(len(responses.columns))]
else:
res_col_num_list = range(len(response_selector))
if input_selector == None:
if input_selector is None:
#saturated model - do all columns
input_selector = [x for x in inp_col_num_list]
for res_col_num in response_selector:
Expand Down Expand Up @@ -540,9 +533,9 @@ def tune_model(inputs,
"""Wrapper to calulate_R2_and_Q2_for_models to make life easy
It does both scaled and unscaled models
assumes you want an unscaled model for ease of plotting
and a scaled model coefficients for ease of choosing"""#
and a scaled model coefficients for ease of choosing"""

## scaled model, use this for picking your coefficients
# scaled model, use this for picking your coefficients
this_model, R2, temp_tuple, selected_input_terms = calulate_R2_and_Q2_for_models(
inputs,
responses,
Expand All @@ -553,7 +546,7 @@ def tune_model(inputs,
do_scaling_here=True)
scaled_model, predictions, ground_truth, coeffs, R2s, R2, Q2 = temp_tuple

## unscaled model, use this for picking your coefficients
# unscaled model, use this for picking your coefficients
#this_model, R2, temp_tuple, selected_input_terms = calulate_R2_and_Q2_for_models(
# inputs,
# responses[['Profit']],
Expand Down Expand Up @@ -658,7 +651,7 @@ def autotune_model(inputs,
except:
if do_hierarchical:
print(
f"Error: Heirarchical model missing lower level terms!!!!"
"Error: Heirarchical model missing lower level terms!!!!"
)
print(dependency_dict)
# Handy shortcut - since the empty set is considered false,
Expand Down Expand Up @@ -708,7 +701,7 @@ def autotune_model(inputs,

for idx, error_value in insignificant_terms:
# If it has dependents, and you're doing an heirarchical model skip it
if do_hierarchical == True:
if do_hierarchical:
if dependency_dict[idx]:
continue
print(
Expand Down Expand Up @@ -964,9 +957,9 @@ def add_higher_order_terms(inputs,
inputs: the input/feature/variable array with data
add_squares=True : whether to add square terms, e.g. x_1^2, x_2^2
add_interactions=True: whether to add interaction terms, x_1*x_2, etc
column_list=[]: to select only a subset of columns, input a column list here
column_list=[]: to select only a subset of columns, input a column list here
Currently does not go above power of 2
returns saturated array and a list of which inputs created which column"""

sat_inputs = copy.deepcopy(inputs)
Expand All @@ -978,7 +971,7 @@ def add_higher_order_terms(inputs,

source_list = [x for x in column_list]

if add_squares == True:
if add_squares:
if verbose:
print('Adding square terms:')
for i in range(len(column_list)):
Expand All @@ -989,7 +982,7 @@ def add_higher_order_terms(inputs,
print(new_name)
sat_inputs[new_name] = inputs[input_name] * inputs[input_name]

if add_interactions == True:
if add_interactions:
if verbose:
print('Adding interaction terms:')
for i in range(len(column_list)):
Expand All @@ -1011,8 +1004,8 @@ def plot_training(R2_over_opt, Q2_over_opt, n_terms_over_opt):
n_terms_over_opt
R2_over_opt: list of R2 over optimisation
Q2_over_opt: list of Q2 over optimisation
n_terms_over_opt: running number of terms
n_terms_over_opt: running number of terms
"""
ax = plt.axes()
x_data = range(len(R2_over_opt))
Expand All @@ -1033,7 +1026,7 @@ def plot_training(R2_over_opt, Q2_over_opt, n_terms_over_opt):
def predict_from_model(model, inputs, input_selector):
"""Reorgs the inputs and does a prediction
model = the model to use
inputs = the saturated inputs
inputs = the saturated inputs
input_selector = the subset of inputs the model is using
"""
list_of_terms = [inputs.columns[x] for x in input_selector]
Expand Down

0 comments on commit b8138da

Please sign in to comment.