search.py

import torch
import numpy as np

device = torch.device("cuda:0" if (torch.cuda.is_available()) else "cpu") 

def decrease(bestmodel):
  '''
      Code for decreasing constant values of inputs (scattering power) to "scan" all possible predicitons of 
      particle dimensions' sizes. Negative values of scatttering power are present as inputs
      since despite the fact that model is not aware of scttering physics for all particles,
      it can still produce efficient predictions for dimension sizes.
  '''
  bestmodel.eval()
  expected = torch.ones(1,5)
  threshold = 9
  j=0
  vals_arr = []
  threshold_arr = []

  while   threshold> -2: 
    vals = expected
    threshold = (0.1-0.0025*j)
      

    true_x_np = threshold *np.ones((1,151))
    true_x = torch.from_numpy( true_x_np.reshape(1,151) ).float()
    expected = bestmodel(true_x.to(device)).view(1,-1)
    vals_arr.append(expected.cpu().data.numpy())
    threshold_arr.append(threshold)
    

  vals_arr = np.array(vals_arr)
  vals_arr = vals_arr.reshape(vals_arr.shape[0],-1)

  threshold_arr = np.array(threshold_arr)

  print(vals_arr.shape)
  print(threshold_arr.shape)
  return vals_arr, threshold_arr 


def extract_positive(vals_arr, threshold_arr): 
  val_temp = []
  threshold_temp= []  
  counter = 0;

  main_val_matlab = []
  main_tr_matlab = []
  for i,row in enumerate(vals_arr):
      if (row<1).any() == False:

        val_temp.append(row)
        threshold_temp.append( threshold_arr[i])
        counter +=1


  print(counter)
  val_temp = np.array(val_temp)
  val_temp.shape

  threshold_temp = np.array(threshold_temp)
  threshold_temp.shape
  
  return  val_temp, threshold_temp