Dee Learning on Rolled Out Pad PCA
rm(list=ls())
load("pid.rdata")
load("rollout_pca")
library(keras)
use_condaenv('r-tensorflow')
pid <- to_categorical(pid)
rm(model.a)
## Warning in rm(model.a): object 'model.a' not found
model.a <- keras_model_sequential()
model.a %>%
layer_dense(units = 32,activation = "relu",input_shape = ncol(pca.dat)) %>%
layer_dense(units = 32,activation = "relu") %>%
layer_dense(units=2,activation = "softmax")
model.a %>% compile(
optimizer=optimizer_adam(),
loss="binary_crossentropy",
metrics=c("accuracy")
)
model.a
## Model
## ___________________________________________________________________________
## Layer (type) Output Shape Param #
## ===========================================================================
## dense_1 (Dense) (None, 32) 5888
## ___________________________________________________________________________
## dense_2 (Dense) (None, 32) 1056
## ___________________________________________________________________________
## dense_3 (Dense) (None, 2) 66
## ===========================================================================
## Total params: 7,010
## Trainable params: 7,010
## Non-trainable params: 0
## ___________________________________________________________________________
model.a %>% fit(pca.dat,pid,epochs=50,validation_split=0.3,batch_size=1000)
model.a %>% save_model_hdf5("/Users/gerhard/Msc-thesis/NEW/Rmodel_a.h5")
rm(model.b)
## Warning in rm(model.b): object 'model.b' not found
model.b <- keras_model_sequential()
model.b %>%
layer_dense(units = 256,activation = "relu",input_shape = ncol(pca.dat),regularizer_l2()) %>%
layer_dense(units = 256,activation = "relu",regularizer_l2()) %>%
layer_dense(units=2,activation = "softmax")
model.b %>% compile(
optimizer=optimizer_adam(),
loss="binary_crossentropy",
metrics=c("accuracy")
)
model.b
## Model
## ___________________________________________________________________________
## Layer (type) Output Shape Param #
## ===========================================================================
## dense_4 (Dense) (None, 256) 47104
## ___________________________________________________________________________
## dense_5 (Dense) (None, 256) 65792
## ___________________________________________________________________________
## dense_6 (Dense) (None, 2) 514
## ===========================================================================
## Total params: 113,410
## Trainable params: 113,410
## Non-trainable params: 0
## ___________________________________________________________________________
model.b %>% fit(pca.dat,pid,epochs=50,validation_split=0.3,batch_size=1000)
model.b %>% save_model_hdf5("/Users/gerhard/Msc-thesis/NEW/Rmodel_b.h5")
test <- predict_classes(model.b,pca.dat)
all(test==0)
rm(list=ls())
load("pca.dat2")
load("pid.rdata")
library(keras)
use_condaenv('r-tensorflow')
pid <- to_categorical(pid)
rm(model.c)
## Warning in rm(model.c): object 'model.c' not found
model.c <- keras_model_sequential()
model.c %>%
layer_dense(units = 32,activation = "relu",input_shape = ncol(pca.dat2)) %>%
layer_dense(units = 32,activation = "relu") %>%
layer_dense(units=2,activation = "softmax")
model.c %>% compile(
optimizer=optimizer_adam(),
loss="binary_crossentropy",
metrics=c("accuracy")
)
model.c
## Model
## ___________________________________________________________________________
## Layer (type) Output Shape Param #
## ===========================================================================
## dense_7 (Dense) (None, 32) 544
## ___________________________________________________________________________
## dense_8 (Dense) (None, 32) 1056
## ___________________________________________________________________________
## dense_9 (Dense) (None, 2) 66
## ===========================================================================
## Total params: 1,666
## Trainable params: 1,666
## Non-trainable params: 0
## ___________________________________________________________________________
model.c %>% fit(pca.dat2,pid,epochs=50,validation_split=0.3,batch_size=1000)
model.c %>% save_model_hdf5("/Users/gerhard/Msc-thesis/NEW/Rmodel_c.h5")
rm(model.d)
## Warning in rm(model.d): object 'model.d' not found
model.d <- keras_model_sequential()
model.d %>%
layer_dense(units = 32,activation = "relu",input_shape = ncol(pca.dat2)) %>%
layer_dense(units = 32,activation = "relu") %>%
layer_dense(units = 32,activation = "relu") %>%
layer_dense(units = 32,activation = "relu") %>%
layer_dense(units = 32,activation = "relu") %>%
layer_dense(units = 32,activation = "relu") %>%
layer_dense(units = 32,activation = "relu") %>%
layer_dense(units = 32,activation = "relu") %>%
layer_dense(units = 32,activation = "relu") %>%
layer_dense(units = 32,activation = "relu") %>%
layer_dense(units=2,activation = "softmax")
model.d %>% compile(
optimizer=optimizer_adam(),
loss="binary_crossentropy",
metrics=c("accuracy")
)
model.d
## Model
## ___________________________________________________________________________
## Layer (type) Output Shape Param #
## ===========================================================================
## dense_10 (Dense) (None, 32) 544
## ___________________________________________________________________________
## dense_11 (Dense) (None, 32) 1056
## ___________________________________________________________________________
## dense_12 (Dense) (None, 32) 1056
## ___________________________________________________________________________
## dense_13 (Dense) (None, 32) 1056
## ___________________________________________________________________________
## dense_14 (Dense) (None, 32) 1056
## ___________________________________________________________________________
## dense_15 (Dense) (None, 32) 1056
## ___________________________________________________________________________
## dense_16 (Dense) (None, 32) 1056
## ___________________________________________________________________________
## dense_17 (Dense) (None, 32) 1056
## ___________________________________________________________________________
## dense_18 (Dense) (None, 32) 1056
## ___________________________________________________________________________
## dense_19 (Dense) (None, 32) 1056
## ___________________________________________________________________________
## dense_20 (Dense) (None, 2) 66
## ===========================================================================
## Total params: 10,114
## Trainable params: 10,114
## Non-trainable params: 0
## ___________________________________________________________________________
model.d %>% fit(pca.dat2,pid,epochs=50,validation_split=0.3,batch_size=1000)
model.d %>% save_model_hdf5("/Users/gerhard/Msc-thesis/NEW/Rmodel_d.h5")
rm(model.e)
## Warning in rm(model.e): object 'model.e' not found
model.e <- keras_model_sequential()
model.e %>%
layer_dense(units = 512,activation = "relu",input_shape = ncol(pca.dat2)) %>%
layer_dense(units = 512,activation = "relu") %>%
layer_dense(units = 32,activation = "relu") %>%
layer_dense(units = 32,activation = "relu") %>%
layer_dense(units = 32,activation = "relu") %>%
layer_dense(units = 32,activation = "relu") %>%
layer_dense(units = 32,activation = "relu") %>%
layer_dense(units = 32,activation = "relu") %>%
layer_dense(units = 32,activation = "relu") %>%
layer_dense(units = 32,activation = "relu") %>%
layer_dense(units=2,activation = "softmax")
model.e %>% compile(
optimizer=optimizer_adam(),
loss="binary_crossentropy",
metrics=c("accuracy")
)
model.e
## Model
## ___________________________________________________________________________
## Layer (type) Output Shape Param #
## ===========================================================================
## dense_21 (Dense) (None, 512) 8704
## ___________________________________________________________________________
## dense_22 (Dense) (None, 512) 262656
## ___________________________________________________________________________
## dense_23 (Dense) (None, 32) 16416
## ___________________________________________________________________________
## dense_24 (Dense) (None, 32) 1056
## ___________________________________________________________________________
## dense_25 (Dense) (None, 32) 1056
## ___________________________________________________________________________
## dense_26 (Dense) (None, 32) 1056
## ___________________________________________________________________________
## dense_27 (Dense) (None, 32) 1056
## ___________________________________________________________________________
## dense_28 (Dense) (None, 32) 1056
## ___________________________________________________________________________
## dense_29 (Dense) (None, 32) 1056
## ___________________________________________________________________________
## dense_30 (Dense) (None, 32) 1056
## ___________________________________________________________________________
## dense_31 (Dense) (None, 2) 66
## ===========================================================================
## Total params: 295,234
## Trainable params: 295,234
## Non-trainable params: 0
## ___________________________________________________________________________
model.e %>% fit(pca.dat2,pid,epochs=50,validation_split=0.3,batch_size=1000)
model.e %>% save_model_hdf5("/Users/gerhard/Msc-thesis/NEW/Rmodel_e.h5")
