$1)$ Complete the following program.

# Small LSTM Network to Generate Text for Alice in Wonderland

import sys

import numpy

from keras.models import Sequential

from keras.layers import Dense, Dropout, LSTM

from keras.callbacks import ModelCheckpoint

#from keras.utils import np$\_$utils

from keras.utils import to$\_$categorical

# load ascii text and covert to lowercase

filename $=$ "wonderland.txt$"$

raw$\_$text $=$ open$($filename$).$read$()$

raw$\_$text $=$ raw$\_$text.lower$()$

# create mapping of unique chars to integers

chars $=$ sorted$($list$($set$($raw$\_$text$)))$

char$\_$to$\_$int $=$ dict$(($c$,$ i$)$ for i$,$ c in enumerate$($chars$))$

$$

Create int$\_$to$\_$int that changes integer to char.

YOUR WORK HERE $(2$ pts$)$

$$

# summarize the loaded data

n$\_$chars $=$ len$($raw$\_$text$)$

n$\_$vocab $=$ len$($chars$)$

print$("$Total Characters: $",$ n$\_$chars$)$

print$("$Total Vocab: $",$ n$\_$vocab$)$

# prepare the dataset of input to output pairs encoded as integers

seq$\_$length $=100$

dataX $=[]$

dataY $=[]$

for i in range$(0,$ n$\_$chars $-$ seq$\_$length, $1)$:

seq$\_$in $=$ raw$\_$text$[$i:i $+$ seq$\_$length$]$

seq$\_$out $=$ raw$\_$text$[$i $+$ seq$\_$length$]$

dataX.append$([$char$\_$to$\_$int$[$char$]$ for char in seq$\_$in$])$

dataY.append$($char$\_$to$\_$int$[$seq$\_$out$])$

n$\_$patterns $=$ len$($dataX$)$

print$("$Total Patterns: $",$ n$\_$patterns$)$

# reshape X to be $[$samples$,$ time steps, features$]$

X $=$ numpy.reshape$($dataX$,($n$\_$patterns, OOO, O$))$

$$

Complete OOO, O

YOUR WORK HERE $(2$ pts$)$

$$

# normalize

X $=$ X $/$ float$($n$\_$vocab$)$

# one hot encode the output variable

y $=$ to$\_$categorical$($dataY$)$

# define the LSTM model

model $=$ Sequential$()$

model.add$($LSTM$(256,$ input$\_$shape$=($X$.$shape$[1],$ X$.$shape$[2])))$

model.add$($Dropout$(0\mathrm{.}2))$

model.add$($Dense$($y$.$shape$[1],$ activation$=$'softmax'$))$

model.compile$($loss$=$'categorical$\_$crossentropy', optimizer$=$'adam'$)$

# define the checkpoint

filepath$=$"weights$-$improvement$-\{$epoch:$02$d$\}-\{$loss:$\mathrm{.}4$f$\}.$keras"

checkpoint $=$ ModelCheckpoint$($filepath$,$ monitor$=$'loss', verbose$=1,$ save$\_$best$\_$only$=$True, mode$=$'min'$)$

callbacks$\_$list $=[$checkpoint$]$

# fit the model

model.fit$($X$,$ y$,$ epochs$=5,$ batch$\_$size$=128,$ callbacks$=$callbacks$\_$list$)$

# pick a random seed

start $=$ numpy.random.randint$(0,$ len$($dataX$)-1)$

pattern $=$ dataX$[$start$]$

print$("$Seed:$")$

#print$("\backslash "",\text{'}\text{'}.$join$([$int$\_$to$\_$char$[$value$]$ for value in pattern$]),"\backslash "")$

# generate characters

for i in range$(100)$:

x $=$ numpy.reshape$($pattern$,(1,$ len$($pattern$),1))$

x $=$ x $/$ float$($n$\_$vocab$)$

prediction $=$ model.predict$($x$,$ verbose$=0)$

index $=$ numpy.argmax$($prediction$)$

result $=$ int$\_$to$\_$char$[$index$]$

seq$\_$in $=[$int$\_$to$\_$char$[$value$]$ for value in pattern$]$

sys$.$stdout.write$($result$)$

pattern.append$($index$)$

pattern $=$ pattern$[1$:len$($pattern$)]$

print$("$

Done."$)$

$2)[3$ pts$]$ increase epochs and see the difference in generated sentences.

$3)[3$ pts$]$ Try some other methods to improve the performance of the program. $($e$.$g$.,$ adding layers, dropout, $...$ etc$)$ Explain the results.