emuccino

#initialize our generators; specifying data directories, batch size, and dimension threshold

train_image_directory = 'imagenette2/train'
test_image_directory = 'imagenette2/val'

n_classes = 10
batch_size = 16
max_dimension = 512

#create generators for training and generating

emuccino

import numpy as np
import pandas as pd

#import loan dataset
df = pd.read_csv('loan.csv').dropna(axis=1,how='any')

#convert loan grades to numerical values
df['sub_grade'] = df['sub_grade'].str.slice(start=1).astype(int)
grade_dict = {k:i for i,k in enumerate(['A', 'B', 'C', 'D', 'E', 'F', 'G'])}
term_dict = {k:i for i,k in enumerate(['36 months', '60 months'])}

emuccino

from sklearn.preprocessing import MinMaxScaler

#specify target label
target = 'loan_condition'

data = [name for name in df.columns if name != target]
numeric_data = []
string_data = []
tokenizers = {}

emuccino

from tensorflow.keras.models import Model
from tensorflow.keras.layers import Input,Dense,BatchNormalization,Concatenate,GaussianNoise
from tensorflow.keras.optimizers import Nadam

#define latent dimension size
latent_dim = int(np.ceil(np.log(len(train_df)*len(data))))

#function for building generator network
def compile_generator():
    

emuccino

from tensorflow.keras.utils import to_categorical   

#function for generating latent samples for synthetic data for generator training
def generate_latent_samples(n):
    
    #generate latent vectors with balanced targets
    x = {'latent': np.random.normal(size=(n, latent_dim)),
         'target': to_categorical(np.hstack([np.array([_x for _ in range(n//2)]) for _x in range(2)]),2)}
    
    #outputs indicating postive discirmination (target value)

emuccino

import itertools
import matplotlib.pyplot as plt

def train_gan(n_epochs,n_batch,n_plot,n_eval):
    
    #discriminator/generator training logs
    disc_loss_hist = []
    gen_loss_hist = []

    for epoch in range(n_epochs):

emuccino

#function for building classifier (very similar to the discriminator)
def compile_classifier():
    
    inputs = {}
    
    numeric_nets = []
    string_nets = []

    for name in numeric_data:
        numeric_input = Input(shape=(1,),name=name)

emuccino

from tensorflow.keras.datasets import mnist

#load mnist data
(x_train,y_train),(x_test,y_test) = mnist.load_data()

#reshape and scale input image data
x_train, x_test = (x.reshape(*x.shape,1)/x.max() for x in (x_train,x_test))

emuccino

from tensorflow.keras.models import Model
from tensorflow.keras.layers import Input, Conv2D, Dense, GlobalAvgPool2D, GlobalMaxPool2D, Concatenate, BatchNormalization
from tensorflow.keras.losses import SparseCategoricalCrossentropy

#function compiling device model
def compile_device_model(input_shape=None,n_filters=None,name=None,offload=False):
    outputs = {}
    
    inputs = Input(shape=input_shape)
    net = inputs

emuccino

outputs = {}

#complete model input
inputs = Input(shape=x_train.shape[1:])
net = {'offload':inputs}

#stack all 3 device models together
for device in device_names:    
    net = device_models[device](net['offload'])
    outputs[device] = net[device+'_outputs']