symbolic-regression/Code/S_separability.py
Silviu Marian Udrescu 13148a5c6a
Add files via upload
2020-04-29 13:41:52 -04:00

384 lines
14 KiB
Python

from __future__ import print_function
import torch
import os
import torch.nn as nn
import torch.nn.functional as F
import torch.optim as optim
import pandas as pd
import numpy as np
import torch
from torch.utils import data
import pickle
from torch.optim.lr_scheduler import CosineAnnealingLR
from matplotlib import pyplot as plt
from itertools import combinations
import time
is_cuda = torch.cuda.is_available()
class SimpleNet(nn.Module):
def __init__(self, ni):
super().__init__()
self.linear1 = nn.Linear(ni, 128)
self.bn1 = nn.BatchNorm1d(128)
self.linear2 = nn.Linear(128, 128)
self.bn2 = nn.BatchNorm1d(128)
self.linear3 = nn.Linear(128, 64)
self.bn3 = nn.BatchNorm1d(64)
self.linear4 = nn.Linear(64,64)
self.bn4 = nn.BatchNorm1d(64)
self.linear5 = nn.Linear(64,1)
def forward(self, x):
x = F.tanh(self.bn1(self.linear1(x)))
x = F.tanh(self.bn2(self.linear2(x)))
x = F.tanh(self.bn3(self.linear3(x)))
x = F.tanh(self.bn4(self.linear4(x)))
x = self.linear5(x)
return x
def rmse_loss(pred, targ):
denom = targ**2
denom = torch.sqrt(denom.sum()/len(denom))
return torch.sqrt(F.mse_loss(pred, targ))/denom
def check_separability_plus(pathdir, filename):
try:
pathdir_weights = "results/NN_trained_models/models/"
# load the data
n_variables = np.loadtxt(pathdir+filename, dtype='str').shape[1]-1
variables = np.loadtxt(pathdir+filename, usecols=(0,))
if n_variables==1:
print(filename, "just one variable for ADD")
# if there is just one variable you have nothing to separate
return (-1,-1,-1)
else:
for j in range(1,n_variables):
v = np.loadtxt(pathdir+filename, usecols=(j,))
variables = np.column_stack((variables,v))
f_dependent = np.loadtxt(pathdir+filename, usecols=(n_variables,))
f_dependent = np.reshape(f_dependent,(len(f_dependent),1))
factors = torch.from_numpy(variables)
if is_cuda:
factors = factors.cuda()
else:
factors = factors
factors = factors.float()
product = torch.from_numpy(f_dependent)
if is_cuda:
product = product.cuda()
else:
product = product
product = product.float()
# load the trained model and put it in evaluation mode
if is_cuda:
model = SimpleNet(n_variables).cuda()
else:
model = SimpleNet(n_variables)
model.load_state_dict(torch.load(pathdir_weights+filename+".h5"))
model.eval()
# make some variables at the time equal to the median of factors
models_one = []
models_rest = []
with torch.no_grad():
fact_vary = factors.clone()
for k in range(len(factors[0])):
fact_vary[:,k] = torch.full((len(factors),),torch.median(factors[:,k]))
# loop through all indices combinations
var_indices_list = np.arange(0,n_variables,1)
min_error = 1000
best_i = []
best_j = []
for i in range(1,n_variables):
c = combinations(var_indices_list, i)
for j in c:
fact_vary_one = factors.clone()
fact_vary_rest = factors.clone()
rest_indx = list(filter(lambda x: x not in j, var_indices_list))
for t1 in rest_indx:
fact_vary_one[:,t1] = torch.full((len(factors),),torch.median(factors[:,t1]))
for t2 in j:
fact_vary_rest[:,t2] = torch.full((len(factors),),torch.median(factors[:,t2]))
# check if the equation is separable
sm = model(fact_vary_one)+model(fact_vary_rest)
#error = torch.sqrt(torch.mean((product-sm+model(fact_vary))**2))/torch.sqrt(torch.mean(product**2))
error = 2*torch.median(abs(product-sm+model(fact_vary)))
if error<min_error:
min_error = error
best_i = j
best_j = rest_indx
return min_error, best_i, best_j
except Exception as e:
print(e)
return (-1,-1,-1)
def do_separability_plus(pathdir, filename, list_i,list_j):
try:
pathdir_weights = "results/NN_trained_models/models/"
# load the data
n_variables = np.loadtxt(pathdir+filename, dtype='str').shape[1]-1
variables = np.loadtxt(pathdir+filename, usecols=(0,))
if n_variables==1:
print(filename, "just one variable for ADD")
# if there is just one variable you have nothing to separate
return (-1,-1,-1)
else:
for j in range(1,n_variables):
v = np.loadtxt(pathdir+filename, usecols=(j,))
variables = np.column_stack((variables,v))
f_dependent = np.loadtxt(pathdir+filename, usecols=(n_variables,))
f_dependent = np.reshape(f_dependent,(len(f_dependent),1))
factors = torch.from_numpy(variables)
if is_cuda:
factors = factors.cuda()
else:
factors = factors
factors = factors.float()
product = torch.from_numpy(f_dependent)
if is_cuda:
product = product.cuda()
else:
product = product
product = product.float()
# load the trained model and put it in evaluation mode
if is_cuda:
model = SimpleNet(n_variables).cuda()
else:
model = SimpleNet(n_variables)
model.load_state_dict(torch.load(pathdir_weights+filename+".h5"))
model.eval()
# make some variables at the time equal to the median of factors
models_one = []
models_rest = []
fact_vary = factors.clone()
for k in range(len(factors[0])):
fact_vary[:,k] = torch.full((len(factors),),torch.median(factors[:,k]))
fact_vary_one = factors.clone()
fact_vary_rest = factors.clone()
for t1 in list_j:
fact_vary_one[:,t1] = torch.full((len(factors),),torch.median(factors[:,t1]))
for t2 in list_i:
fact_vary_rest[:,t2] = torch.full((len(factors),),torch.median(factors[:,t2]))
with torch.no_grad():
str1 = filename+"-add_a"
str2 = filename+"-add_b"
# save the first half
data_sep_1 = variables
data_sep_1 = np.delete(data_sep_1,list_j,axis=1)
data_sep_1 = np.column_stack((data_sep_1,model(fact_vary_one).cpu()))
# save the second half
data_sep_2 = variables
data_sep_2 = np.delete(data_sep_2,list_i,axis=1)
data_sep_2 = np.column_stack((data_sep_2,model(fact_vary_rest).cpu()-model(fact_vary).cpu()))
try:
os.mkdir("results/separable_add/")
except:
pass
np.savetxt("results/separable_add/"+str1,data_sep_1)
np.savetxt("results/separable_add/"+str2,data_sep_2)
# if it is separable, return the 2 new files created and the index of the column with the separable variable
return ("results/separable_add/",str1,"results/separable_add/",str2)
except Exception as e:
print(e)
return (-1,-1)
def check_separability_multiply(pathdir, filename):
try:
pathdir_weights = "results/NN_trained_models/models/"
# load the data
n_variables = np.loadtxt(pathdir+filename, dtype='str').shape[1]-1
variables = np.loadtxt(pathdir+filename, usecols=(0,))
if n_variables==1:
print(filename, "just one variable for ADD")
# if there is just one variable you have nothing to separate
return (-1,-1,-1)
else:
for j in range(1,n_variables):
v = np.loadtxt(pathdir+filename, usecols=(j,))
variables = np.column_stack((variables,v))
f_dependent = np.loadtxt(pathdir+filename, usecols=(n_variables,))
# Pick only data which is close enough to the maximum value (5 times less or higher)
max_output = np.max(abs(f_dependent))
use_idx = np.where(abs(f_dependent)>=max_output/5)
f_dependent = f_dependent[use_idx]
f_dependent = np.reshape(f_dependent,(len(f_dependent),1))
variables = variables[use_idx]
factors = torch.from_numpy(variables)
if is_cuda:
factors = factors.cuda()
else:
factors = factors
factors = factors.float()
product = torch.from_numpy(f_dependent)
if is_cuda:
product = product.cuda()
else:
product = product
product = product.float()
# load the trained model and put it in evaluation mode
if is_cuda:
model = SimpleNet(n_variables).cuda()
else:
model = SimpleNet(n_variables)
model.load_state_dict(torch.load(pathdir_weights+filename+".h5"))
model.eval()
# make some variables at the time equal to the median of factors
models_one = []
models_rest = []
with torch.no_grad():
fact_vary = factors.clone()
for k in range(len(factors[0])):
fact_vary[:,k] = torch.full((len(factors),),torch.median(factors[:,k]))
# loop through all indices combinations
var_indices_list = np.arange(0,n_variables,1)
min_error = 1000
best_i = []
best_j = []
for i in range(1,n_variables):
c = combinations(var_indices_list, i)
for j in c:
fact_vary_one = factors.clone()
fact_vary_rest = factors.clone()
rest_indx = list(filter(lambda x: x not in j, var_indices_list))
for t1 in rest_indx:
fact_vary_one[:,t1] = torch.full((len(factors),),torch.median(factors[:,t1]))
for t2 in j:
fact_vary_rest[:,t2] = torch.full((len(factors),),torch.median(factors[:,t2]))
# check if the equation is separable
pd = model(fact_vary_one)*model(fact_vary_rest)
#error = torch.sqrt(torch.mean((product-pd/model(fact_vary))**2))/torch.sqrt(torch.mean(product**2))
error = 2*torch.median(abs(product-pd/model(fact_vary)))
if error<min_error:
min_error = error
best_i = j
best_j = rest_indx
return min_error, best_i, best_j
except Exception as e:
print(e)
return (-1,-1,-1)
def do_separability_multiply(pathdir, filename, list_i,list_j):
try:
pathdir_weights = "results/NN_trained_models/models/"
# load the data
n_variables = np.loadtxt(pathdir+filename, dtype='str').shape[1]-1
variables = np.loadtxt(pathdir+filename, usecols=(0,))
if n_variables==1:
print(filename, "just one variable for ADD")
# if there is just one variable you have nothing to separate
return (-1,-1,-1)
else:
for j in range(1,n_variables):
v = np.loadtxt(pathdir+filename, usecols=(j,))
variables = np.column_stack((variables,v))
f_dependent = np.loadtxt(pathdir+filename, usecols=(n_variables,))
f_dependent = np.reshape(f_dependent,(len(f_dependent),1))
factors = torch.from_numpy(variables)
if is_cuda:
factors = factors.cuda()
else:
factors = factors
factors = factors.float()
product = torch.from_numpy(f_dependent)
if is_cuda:
product = product.cuda()
else:
product = product
product = product.float()
# load the trained model and put it in evaluation mode
if is_cuda:
model = SimpleNet(n_variables).cuda()
else:
model = SimpleNet(n_variables)
model.load_state_dict(torch.load(pathdir_weights+filename+".h5"))
model.eval()
# make some variables at the time equal to the median of factors
models_one = []
models_rest = []
fact_vary = factors.clone()
for k in range(len(factors[0])):
fact_vary[:,k] = torch.full((len(factors),),torch.median(factors[:,k]))
fact_vary_one = factors.clone()
fact_vary_rest = factors.clone()
for t1 in list_j:
fact_vary_one[:,t1] = torch.full((len(factors),),torch.median(factors[:,t1]))
for t2 in list_i:
fact_vary_rest[:,t2] = torch.full((len(factors),),torch.median(factors[:,t2]))
with torch.no_grad():
str1 = filename+"-mult_a"
str2 = filename+"-mult_b"
# save the first half
data_sep_1 = variables
data_sep_1 = np.delete(data_sep_1,list_j,axis=1)
data_sep_1 = np.column_stack((data_sep_1,model(fact_vary_one).cpu()))
# save the second half
data_sep_2 = variables
data_sep_2 = np.delete(data_sep_2,list_i,axis=1)
data_sep_2 = np.column_stack((data_sep_2,model(fact_vary_rest).cpu()/model(fact_vary).cpu()))
try:
os.mkdir("results/separable_mult/")
except:
pass
np.savetxt("results/separable_mult/"+str1,data_sep_1)
np.savetxt("results/separable_mult/"+str2,data_sep_2)
# if it is separable, return the 2 new files created and the index of the column with the separable variable
return ("results/separable_mult/",str1,"results/separable_mult/",str2)
except Exception as e:
print(e)
return (-1,-1)