def closest_coords(target_file): data = read_data(target_file) closest = (0, 0, float('inf')) for idx1, coord1 in enumerate(data): for idx2, coord2 in enumerate(data): distance = euclidean_dist(coord1, coord2) if 0 < distance < closest[2]: closest = (idx1, idx2, distance) print(f'{closest[0] + 1}:' + floatlist_to_string(data[closest[0]])) print(f'{closest[1] + 1}:' + floatlist_to_string(data[closest[1]]))
__author__ = 'skao' import data_tools as dt from keras.models import load_model import numpy as np model = load_model('modelD') images, labels = dt.read_data('data/training/training.csv') images, labels = dt.read_data('data/test/test.csv') #========================================== # Analyze Model #========================================== test_batch, label_batch = dt.get_batch(images, labels) results = model.predict(test_batch) dist = [dt.distance(a, b) for a, b in zip(results, label_batch)] #Best and Worst Performing worst = np.argsort(dist)[-16:] best = np.argsort(dist)[:16] dt.create_image_display(test_batch[worst], results[worst], label_batch[worst], False, 2, 4) dt.create_image_display(test_batch[best], results[best], label_batch[best], False, 2, 4) #========================================== # Rotations #========================================== rot_batch, rot_label = dt.get_batch(dt.rotate_images(images, 45, True),
def load_data( self ): path = self.params.save_tmp % ( self.params.data_root, self.params.models_tag, self.n ) self.peptides = data_tools.read_data( self.params.data_path ) self.benchmark = ml_tools.rt_benchmark(self.peptides, 'elude', 'gp', self.n, self.params.nparts, self.params.train_ratio ) self.models, self.kernels = ml_tools.load_rt_models( path )
#!/usr/bin/python import os import data_tools import feature_extraction as fe if __name__=="__main__" : pwd = os.path.dirname(os.path.realpath(__file__)) # Reading peptides and their retention time from file peptides = data_tools.read_data( pwd + '/../Data/20110922_EXQ4_NaNa_SA_YeastEasy_Labelfree_01.rtimes_q_0.001.tsv') # Building different models mgen = fe.model_generator( peptides ) aa_list = mgen.amino_acid_list() elude_model = mgen.get_elude_model() bow_voc = mgen.get_bow_voc( 2 ) # K = number of letters in each word peptide = peptides[0] print "Elude descriptor is" print peptide.elude_descriptor( elude_model ) print "Bow Descriptor is" print peptide.bow_descriptor( bow_voc )
import pickle import GPy if __name__ == "__main__": n = 1000 models_path = "/Users/heydar/Stuff/tmp/gprt/models_ntrain_%d.pk" % (n) with open(models_path, "r") as ff: models = pickle.load(ff)[0] ff.close() print len(models) raw_input() peptides = data_tools.read_data() # duplicated_message = data_tools.checked_duplicated(peptides) # print duplicated_message bench = ml_tools.rt_benchmark(peptides, "elude", "gp", 100, 5) fmat = [] mmat = [] dmat = [] for i in range(bench.parts.nfolds): print i model = bench.train_model(i) f, m, d = bench.test_sorted(i, model) fmat.append(f)
__author__ = 'skao' import data_tools as dt from keras.models import load_model import numpy as np model = load_model('modelD') images, labels = dt.read_data('data/training/training.csv') images, labels = dt.read_data('data/test/test.csv') #========================================== # Analyze Model #========================================== test_batch, label_batch = dt.get_batch(images, labels) results = model.predict(test_batch) dist = [dt.distance(a,b) for a,b in zip(results, label_batch)] #Best and Worst Performing worst = np.argsort(dist)[-16:] best = np.argsort(dist)[:16] dt.create_image_display(test_batch[worst], results[worst], label_batch[worst], False, 2, 4) dt.create_image_display(test_batch[best], results[best], label_batch[best], False, 2, 4) #========================================== # Rotations #========================================== rot_batch, rot_label = dt.get_batch(dt.rotate_images(images, 45, True), dt.rotate_labels(labels, 45)) rot_results = model.predict(rot_batch)
__author__ = 'skao' import data_tools as dt from keras.layers import Dense, Convolution2D, Input, MaxPooling2D, Flatten, Dropout from keras.models import Model import pandas as pd #========================================== # Reading in data #========================================== images, labels = dt.read_data('data/training/training.csv') #images, labels = dt.get_batch(images, labels, None, True) #images, labels = dt.get_batch(images, labels, 600, False) #for angle in [45,90,135,180,-45,-90,-135]: # images = dt.np.vstack((images,dt.rotate_images(images, angle, reshape_bool=True))) # labels = dt.np.vstack((labels,dt.rotate_labels(labels, angle))) #pd.DataFrame(images).to_csv('data/training/training_images.csv') #pd.DataFrame(labels).to_csv('data/training/training_labels.csv') validate_data, validate_labels = dt.read_data('data/test/test.csv') validate_data, validate_labels = dt.get_batch(validate_data, validate_labels, validate_data.shape[0]) #========================================== # Define Input Shape #========================================== input = Input(shape=(1, 96, 96), dtype='float32') #========================================== # Define Model Architecture #==========================================
#!/usr/bin/python import numpy as np import data_tools import ml_tools import pickle as pk from common import parameters if __name__ == "__main__": params = parameters() peptides = data_tools.read_data( params.data_path ) for n in params.ntrain : print n benchmark = ml_tools.rt_benchmark( peptides, 'elude', 'gp', n , params.nparts, params.train_ratio ) models = ml_tools.single_train_gp( benchmark ) save_path = params.save_tmp % ( params.data_root, params.models_tag, n ) with open( save_path, 'w' ) as ff : pk.dump( [ models ], ff ) ff.close() models = None