-
Notifications
You must be signed in to change notification settings - Fork 8
/
wavKAN.py
115 lines (104 loc) · 4.7 KB
/
wavKAN.py
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
import torch
import torch.nn as nn
import torch.nn.functional as F
import torch.optim as optim
import torchvision
import torchvision.transforms as transforms
from torch.utils.data import DataLoader
from tqdm import tqdm
import pandas as pd
from KAN import *
# Defining the wavelet types
#wavelet_types = ['mexican_hat', 'morlet', 'dog', 'meyer', 'shannon', 'bump', etc.] #It can include #all wavelet types
wavelet_types = ['mexican_hat', 'morlet', 'dog']
# Loading MNIST data set
transform = transforms.Compose(
[transforms.ToTensor(), transforms.Normalize((0.5,), (0.5,))]
)
trainset = torchvision.datasets.MNIST(root="./data", train=True, download=True, transform=transform)
valset = torchvision.datasets.MNIST(root="./data", train=False, download=True, transform=transform)
trainloader = DataLoader(trainset, batch_size=64, shuffle=True)
valloader = DataLoader(valset, batch_size=64, shuffle=False)
# Trials and Epochs (epochs per trial)
trials = 5
epochs_per_trial = 50
# Looping over each wavelet type
for wavelet in wavelet_types:
all_train_losses, all_train_accuracies = [], []
all_val_losses, all_val_accuracies = [], []
print(f'Wavelet is {wavelet}')
#For a specified number of trials
for trial in range(trials):
print(f'Trial is {trial}')
# Define model, optimizer, scheduler for each trial
model = KAN([28 * 28, 32, 10], wavelet_type=wavelet)
device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
model.to(device)
optimizer = optim.AdamW(model.parameters(), lr=1e-3, weight_decay=1e-4)
scheduler = optim.lr_scheduler.ExponentialLR(optimizer, gamma=0.9)
criterion = nn.CrossEntropyLoss()
trial_train_losses, trial_val_losses = [], []
trial_train_accuracies, trial_val_accuracies = [], []
#For a specified number of epchs
for epoch in range(epochs_per_trial):
# Training
train_loss, train_correct, train_total = 0.0, 0, 0
model.train()
#for images, labels in tqdm(trainloader):
for images, labels in trainloader:
images = images.view(-1, 28 * 28).to(device)
labels = labels.to(device)
optimizer.zero_grad()
outputs = model(images)
loss = criterion(outputs, labels)
loss.backward()
optimizer.step()
train_loss += loss.item()
_, predicted = torch.max(outputs.data, 1)
train_total += labels.size(0)
train_correct += (predicted == labels).sum().item()
train_loss /= len(trainloader)
train_acc = 100 * train_correct / train_total
trial_train_losses.append(train_loss)
trial_train_accuracies.append(train_acc)
# Validation
val_loss, val_correct, val_total = 0.0, 0, 0
model.eval()
with torch.no_grad():
for images, labels in valloader:
images = images.view(-1, 28 * 28).to(device)
labels = labels.to(device)
outputs = model(images)
loss = criterion(outputs, labels)
val_loss += loss.item()
_, predicted = torch.max(outputs.data, 1)
val_total += labels.size(0)
val_correct += (predicted == labels).sum().item()
val_loss /= len(valloader)
val_acc = 100 * val_correct / val_total
trial_val_losses.append(val_loss)
trial_val_accuracies.append(val_acc)
# Update learning rate
scheduler.step()
#collecting statistics
all_train_losses.append(trial_train_losses)
all_train_accuracies.append(trial_train_accuracies)
all_val_losses.append(trial_val_losses)
all_val_accuracies.append(trial_val_accuracies)
# Average results across trials and write to Excel
avg_train_losses = pd.DataFrame(all_train_losses).mean().tolist()
avg_train_accuracies = pd.DataFrame(all_train_accuracies).mean().tolist()
avg_val_losses = pd.DataFrame(all_val_losses).mean().tolist()
avg_val_accuracies = pd.DataFrame(all_val_accuracies).mean().tolist()
results_df = pd.DataFrame({
'Epoch': range(1, epochs_per_trial + 1),
'Train Loss': avg_train_losses,
'Train Accuracy': avg_train_accuracies,
'Validation Loss': avg_val_losses,
'Validation Accuracy': avg_val_accuracies
})
# Saving the results
# Saving the results to an Excel file named after the wavelet type
file_name = f'{wavelet}_results.xlsx'
results_df.to_excel(file_name, index=False)
print(f"Results saved to {file_name}.")