From 04713d43d1131d3108fef85fc97b9c04b4ea83a1 Mon Sep 17 00:00:00 2001
From: Francois Vieille <vieille.francois@gmail.com>
Date: Thu, 14 Mar 2019 10:46:59 +0100
Subject: [PATCH] variational model

---
 iss/models/VariationalAutoEncoder.py | 72 ++++++++++++++++++++++++++++
 iss/models/__init__.py               |  1 +
 2 files changed, 73 insertions(+)
 create mode 100644 iss/models/VariationalAutoEncoder.py

diff --git a/iss/models/VariationalAutoEncoder.py b/iss/models/VariationalAutoEncoder.py
new file mode 100644
index 0000000..868d2fd
--- /dev/null
+++ b/iss/models/VariationalAutoEncoder.py
@@ -0,0 +1,72 @@
+# -*- coding: utf-8 -*-
+
+from iss.models import AbstractAutoEncoderModel
+from keras.layers import Input, Dense, Conv2D, MaxPooling2D, UpSampling2D, Reshape, Flatten, BatchNormalization, Activation, Lambda
+from keras.optimizers import Adadelta, Adam
+from keras.models import Model
+from keras.losses import binary_crossentropy
+from keras import backend as K
+import numpy as np
+
+class VarAutoEncoder(AbstractAutoEncoderModel):
+
+	def __init__(self, config):
+
+		save_directory = config['save_directory']
+		model_name = config['model_name']
+
+		super().__init__(save_directory, model_name)
+
+		self.activation = config['activation']
+		self.input_shape = (config['input_height'], config['input_width'], config['input_channel'])
+		self.latent_shape = config['latent_shape']
+		self.lr = config['learning_rate']
+		self.build_model()
+
+	def sampling(self, args):
+		z_mean, z_log_var = args
+
+		batch = K.shape(z_mean)[0]
+		dim = K.int_shape(z_mean)[1]
+		epsilon = K.random_normal(shape=(batch, dim))
+		return z_mean + K.exp(0.5 * z_log_var) * epsilon
+
+	def build_model(self):
+		input_shape = self.input_shape
+		latent_shape = self.latent_shape
+
+		picture = Input(shape = input_shape)
+
+		x = Flatten()(picture)
+		x = Dense(1000, activation = 'relu', name = 'enc_1')(x)
+		x = Dense(100, activation = 'relu', name = 'enc_2')(x)
+		z_mean = Dense(latent_shape, name = 'enc_z_mean')(x)
+		z_log_var = Dense(latent_shape, name = 'enc_z_log_var')(x)
+
+
+		z = Lambda(self.sampling, name='enc_z')([z_mean, z_log_var])
+
+		self.encoder_model = Model(picture, [z_mean, z_log_var, z], name = "encoder")
+		
+		latent_input = Input(shape = (latent_shape,), name = "enc_z_sampling")
+		x = Dense(100, activation = 'relu', name = 'dec_1')(latent_input)
+		x = Dense(1000, activation = 'relu', name = 'dec_2')(x)
+		x = Dense(np.prod(input_shape), activation = self.activation)(x)
+		decoded = Reshape((input_shape))(x)
+
+		self.decoder_model = Model(latent_input, decoded, name = "decoder")
+
+		picture_dec = self.decoder_model(self.encoder_model(picture)[2]) 
+		self.model = Model(picture, picture_dec, name = "autoencoder")
+
+		def my_loss(picture, picture_dec):
+
+			xent_loss = K.mean(K.binary_crossentropy(picture, picture_dec), axis = (-1, -2, -3))
+			kl_loss = - 0.5 * K.sum(1 + z_log_var - K.square(z_mean) - K.exp(z_log_var), axis=-1)
+			loss =  K.mean(xent_loss + kl_loss)
+			return loss
+
+		optimizer = Adam(lr = self.lr, beta_1=0.9, beta_2=0.999, epsilon=None, decay=0.0, amsgrad=False)
+		
+		self.model.compile(optimizer = optimizer, loss = my_loss)
+
diff --git a/iss/models/__init__.py b/iss/models/__init__.py
index 027e674..6502272 100644
--- a/iss/models/__init__.py
+++ b/iss/models/__init__.py
@@ -3,3 +3,4 @@ from .AbstractModel import AbstractModel
 from .AbstractModel import AbstractAutoEncoderModel
 from .SimpleConvAutoEncoder import SimpleConvAutoEncoder
 from .SimpleAutoEncoder import SimpleAutoEncoder
+from .VariationalAutoEncoder import VarAutoEncoder