Example 14: Knot supervised
import pandas as pd
import numpy as np
import torch
from kan import *
import copy
device = torch.device('cuda' if torch.cuda.is_available() else 'cpu')
print(device)
seed = 42
torch.manual_seed(seed)
np.random.seed(seed)
# Download data: https://colab.research.google.com/github/deepmind/mathematics_conjectures/blob/main/knot_theory.ipynb#scrollTo=l10N2ZbHu6Ob
df = pd.read_csv("./knot_data.csv")
df.keys()
X = df[df.keys()[1:-1]].to_numpy()
Y = df[['signature']].to_numpy()
# normalize X
X_mean = np.mean(X, axis=0)
X_std = np.std(X, axis=0)
X = (X - X_mean[np.newaxis,:])/X_std[np.newaxis,:]
input_normalier = [X_mean, X_std]
# normalize Y
max_signature = np.max(Y)
min_signature = np.min(Y)
Y = ((Y-min_signature)/2).astype(int)
n_class = int((max_signature-min_signature)/2+1)
output_normalier = [min_signature, 2]
dataset = {}
num = X.shape[0]
n_feature = X.shape[1]
train_ratio = 0.8
train_id_ = np.random.choice(num, int(num*train_ratio), replace=False)
test_id_ = np.array(list(set(range(num))-set(train_id_)))
dtype = torch.get_default_dtype()
dataset['train_input'] = torch.from_numpy(X[train_id_]).type(dtype).to(device)
dataset['train_label'] = torch.from_numpy(Y[train_id_][:,0]).type(torch.long).to(device)
dataset['test_input'] = torch.from_numpy(X[test_id_]).type(dtype).to(device)
dataset['test_label'] = torch.from_numpy(Y[test_id_][:,0]).type(torch.long).to(device)
---------------------------------------------------------------------------
FileNotFoundError Traceback (most recent call last)
/var/folders/6j/b6y80djd4nb5hl73rv3sv8y80000gn/T/ipykernel_75986/3212158569.py in <module>
11
12 # Download data: https://colab.research.google.com/github/deepmind/mathematics_conjectures/blob/main/knot_theory.ipynb#scrollTo=l10N2ZbHu6Ob
---> 13 df = pd.read_csv("./knot_data.csv")
14 df.keys()
15
~/opt/anaconda3/lib/python3.9/site-packages/pandas/util/_decorators.py in wrapper(*args, **kwargs)
309 stacklevel=stacklevel,
310 )
--> 311 return func(*args, **kwargs)
312
313 return wrapper
~/opt/anaconda3/lib/python3.9/site-packages/pandas/io/parsers/readers.py in read_csv(filepath_or_buffer, sep, delimiter, header, names, index_col, usecols, squeeze, prefix, mangle_dupe_cols, dtype, engine, converters, true_values, false_values, skipinitialspace, skiprows, skipfooter, nrows, na_values, keep_default_na, na_filter, verbose, skip_blank_lines, parse_dates, infer_datetime_format, keep_date_col, date_parser, dayfirst, cache_dates, iterator, chunksize, compression, thousands, decimal, lineterminator, quotechar, quoting, doublequote, escapechar, comment, encoding, encoding_errors, dialect, error_bad_lines, warn_bad_lines, on_bad_lines, delim_whitespace, low_memory, memory_map, float_precision, storage_options)
676 kwds.update(kwds_defaults)
677
--> 678 return _read(filepath_or_buffer, kwds)
679
680
~/opt/anaconda3/lib/python3.9/site-packages/pandas/io/parsers/readers.py in _read(filepath_or_buffer, kwds)
573
574 # Create the parser.
--> 575 parser = TextFileReader(filepath_or_buffer, **kwds)
576
577 if chunksize or iterator:
~/opt/anaconda3/lib/python3.9/site-packages/pandas/io/parsers/readers.py in __init__(self, f, engine, **kwds)
930
931 self.handles: IOHandles | None = None
--> 932 self._engine = self._make_engine(f, self.engine)
933
934 def close(self):
~/opt/anaconda3/lib/python3.9/site-packages/pandas/io/parsers/readers.py in _make_engine(self, f, engine)
1214 # "Union[str, PathLike[str], ReadCsvBuffer[bytes], ReadCsvBuffer[str]]"
1215 # , "str", "bool", "Any", "Any", "Any", "Any", "Any"
-> 1216 self.handles = get_handle( # type: ignore[call-overload]
1217 f,
1218 mode,
~/opt/anaconda3/lib/python3.9/site-packages/pandas/io/common.py in get_handle(path_or_buf, mode, encoding, compression, memory_map, is_text, errors, storage_options)
784 if ioargs.encoding and "b" not in ioargs.mode:
785 # Encoding
--> 786 handle = open(
787 handle,
788 ioargs.mode,
FileNotFoundError: [Errno 2] No such file or directory: './knot_data.csv'
def train_acc():
return torch.mean((torch.argmax(model(dataset['train_input']), dim=1) == dataset['train_label']).float())
def test_acc():
return torch.mean((torch.argmax(model(dataset['test_input']), dim=1) == dataset['test_label']).float())
model = KAN(width=[n_feature,1,n_class], grid=5, k=3, seed=seed, device=device)
model.fit(dataset, lamb=0.005, batch=1024, loss_fn = nn.CrossEntropyLoss(), metrics=[train_acc, test_acc], display_metrics=['train_loss', 'reg', 'train_acc', 'test_acc']);
model.plot(scale=1.0, beta=0.2)
n = 17
for i in range(n):
plt.gcf().get_axes()[0].text(1/(2*n)+i/n-0.005,-0.02,df.keys()[1:-1][i], rotation=270, rotation_mode="anchor")
scores = model.feature_score
features = list(df.keys()[1:-1])
y_pos = range(len(features))
plt.bar(y_pos, scores)
plt.xticks(y_pos, features, rotation=90);
plt.ylabel('feature importance')