dekker.one
Group work for a Monash Research Methods course
Move overfitting footnote to it's first occurrence
+13
-14
+13
-14
mini_proj/report/waldo.tex
+13
-14
mini_proj/report/waldo.tex
···
149
149
(binary) tree. Each non-leaf node contain a selection criteria to its
150
150
branches. Every leaf node contains the class that will be assigned to the
151
151
instance if the node is reached. In other training methods, decision trees
152
-
have the tendency to overfit, but in random forest a multitude of decision
153
-
tree is trained with a certain degree of randomness and the mean of these
154
-
trees is used which avoids this problem.
152
+
have the tendency to overfit\footnote{Overfitting occurs when a model learns
153
+
from the data too specifically, and loses its ability to generalise its
154
+
predictions for new data (resulting in loss of prediction accuracy)}, but in
155
+
random forest a multitude of decision tree is trained with a certain degree
156
+
of randomness and the mean of these trees is used which avoids this problem.
155
157
156
158
\subsection{Neural Network Architectures}
157
159
···
233
235
models; requiring training on a dataset of typical images. Each network was
234
236
trained using the preprocessed training dataset and labels, for 25 epochs
235
237
(one forward and backward pass of all data) in batches of 150. The number of
236
-
epochs was chosen to maximise training time and prevent
237
-
overfitting\footnote{Overfitting occurs when a model learns from the data
238
-
too specifically, and loses its ability to generalise its predictions for
239
-
new data (resulting in loss of prediction accuracy)} of the training data,
240
-
given current model parameters. The batch size is the number of images sent
241
-
through each pass of the network. Using the entire dataset would train the
242
-
network quickly, but decrease the network's ability to learn unique features
243
-
from the data. Passing one image at a time may allow the model to learn more
244
-
about each image, however it would also increase the training time and risk
245
-
of overfitting the data. Therefore the batch size was chosen to maintain
246
-
training accuracy while minimising training time.
238
+
epochs was chosen to maximise training time and prevent overfitting of the
239
+
training data, given current model parameters. The batch size is the number
240
+
of images sent through each pass of the network. Using the entire dataset
241
+
would train the network quickly, but decrease the network's ability to learn
242
+
unique features from the data. Passing one image at a time may allow the
243
+
model to learn more about each image, however it would also increase the
244
+
training time and risk of overfitting the data. Therefore the batch size was
245
+
chosen to maintain training accuracy while minimising training time.
247
246
248
247
\subsection{Neural Network Testing}\label{nnTesting}
249
248