Tensorflow
Contents
10.19. Tensorflow¶
Tensorflow is a software library commonly used for developing machine learning applications including detecting objects, recognizing images, and creating sequential models.
Offers: - building and training models with Keras - impressive graphing capabilities - supports powerful add on libraries - good for large projects and complex work flows - share trained models without including the full code
Downsides: - slower than similar programs like PyTorch - only fully supports Python
10.19.1. Background¶
Tensor = multidimensional array, where - n-d tensor = n indicies to access single value of interest - ‘n = 0’: scalar, ‘n = 1’: vector, etc.
Flow - successive execution of an operation through n-d tensor - dataflow graphs - structures describing how data moves through graph/series
10.19.2. Installation and Python Implementation¶
‘pip install tensorflow’
import pandas as pd
from sklearn.model_selection import train_test_split
url = \
"https://raw.githubusercontent.com/statds/ids-s22/main/notes/data/nyc_mv_collisions_202201.csv"
nyc_crash = pd.read_csv(url)
nyc_crash.columns
nyc_crash[nyc_crash['CONTRIBUTING FACTOR VEHICLE 1'] == 'Driver Inattention/Distraction']
nyc_crash['DISTRACTED DRIVING'] = [1 if x == 'Driver Inattention/Distraction' else 0
for x in nyc_crash['CONTRIBUTING FACTOR VEHICLE 1']]
x = pd.get_dummies(nyc_crash.drop(['ZIP CODE', 'LATITUDE', 'LONGITUDE', 'COLLISION_ID', 'LOCATION',
'CRASH DATE', 'CRASH TIME', 'ON STREET NAME', 'CROSS STREET NAME',
'OFF STREET NAME', 'VEHICLE TYPE CODE 1', 'VEHICLE TYPE CODE 2',
'VEHICLE TYPE CODE 3', 'VEHICLE TYPE CODE 4', 'VEHICLE TYPE CODE 5',
'CONTRIBUTING FACTOR VEHICLE 2', 'CONTRIBUTING FACTOR VEHICLE 3',
'CONTRIBUTING FACTOR VEHICLE 4', 'CONTRIBUTING FACTOR VEHICLE 5'], axis = 1))
y = nyc_crash['DISTRACTED DRIVING']
x_train, x_test, y_train, y_test = train_test_split(x, y, test_size = .2)
from tensorflow.keras.models import Sequential, load_model
from tensorflow.keras.layers import Dense
from sklearn.metrics import accuracy_score
# Sequential Models
# - one input tensor to one output tensor
ddriving = Sequential()