From c83464daf4a1f5d3b72f317fe97294cda9f55278 Mon Sep 17 00:00:00 2001
From: corentin <corentin.massala@gmail.com>
Date: Thu, 9 Nov 2023 11:35:44 +0100
Subject: [PATCH] Add of graphs
---
README.md | 40 ++++++++++++++++++++++++++++++++--------
1 file changed, 32 insertions(+), 8 deletions(-)
diff --git a/README.md b/README.md
index 7b59861..014314a 100644
--- a/README.md
+++ b/README.md
@@ -126,7 +126,7 @@ Here is the graph of the accuracy vs K for the whole Cifar dataset with a split
-Here we can conclude that the best K is 9, (if we don't use k = 1) with a performace of 35% of accuracy.
+Here we can conclude that the best K is 9, (if we don't use k = 1) with a performace of 35% of accuracy. (I tried to updtade the graph, but my kernel kept dying at each run so I kept the first version that could execute.)
## Artificial Neural Network
@@ -326,12 +326,11 @@ def run_mlp_training(data_train, labels_train, data_test, labels_test, d_h,learn
d_in = data_train.shape[1]
d_out = 10 #we can hard code it here or len(np.unique(label_train))
- #Random initialisation of weights
- w1 = np.random.randn(d_in, d_h)
- b1 = np.random.randn(1, d_h)
-
- w2 = np.random.randn(d_h, d_out)
- b2 = np.random.randn(1, d_out)
+ #Random initialisation of weights (Xavier initialisation)
+ w1 = np.random.randn(d_in, d_h) / np.sqrt(d_in)
+ b1 = np.zeros((1, d_h))
+ w2 = np.random.randn(d_h, d_out) / np.sqrt(d_h)
+ b2 = np.zeros((1, d_out))
# Train MLP
w1, b1, w2, b2, train_accuracies = train_mlp(w1, b1, w2, b2, data_train, labels_train, learning_rate, num_epoch)
@@ -339,4 +338,29 @@ def run_mlp_training(data_train, labels_train, data_test, labels_test, d_h,learn
# Test MLP
test_accuracy = test_mlp(w1, b1, w2, b2, data_test, labels_test)
return train_accuracies, test_accuracy
-```
\ No newline at end of file
+```
+
+#### 16-
+```rb
+def plot_graph(data_train, labels_train, data_test, labels_test, d_h, learning_rate, num_epoch):
+ # Run MLP training
+ train_accuracies, test_accuracy = run_mlp_training(data_train, labels_train, data_test, labels_test, d_h, learning_rate, num_epoch)
+
+ # Plot and save the learning accuracy graph
+ plt.figure(figsize=(8, 6))
+ epochs = np.arange(1, num_epoch + 1)
+ plt.plot(epochs, train_accuracies, marker='x', color='b', label='Train Accuracy')
+ plt.xlabel('Epochs')
+ plt.ylabel('Accuracy')
+ plt.title('MLP Train Accuracy')
+ plt.legend()
+ plt.grid(True)
+ plt.savefig('image-classification/results/mlp.png')
+ plt.show()
+```
+
+
+
+The accuracy is increasing with each epochs without converging, we could increase the learning rate to speed up the training and inscrease the numbers of epoch to see what would be our maximum accuracy.
+For 100 epochs and a learning rate of 0.1 we got a test accuracy of 0.13.
+For 300 epochs and a learning rate of 0.1 we increased the training accuracy to 0.15991 and we got a test accuracy of 0.155
\ No newline at end of file
--
GitLab