def dim_reduction(self, slash_filter):
scaler = StandardScaler()
scaler.fit(
self.data) # Fit Data with StandardScaler() (get mean and stddev)
self.data = scaler.transform(
self.data) # Transform original data set based on the .fita
pca = decomposition.PCA(n_components=2)
pca.fit(self.data)
self.data = pca.transform(self.data)
# Renormalise the data so that the values of our pca sit nicely on a grid.
self.pca_data = norm_np(self.data)
self.pca_data_transposed = self.pca_data.transpose()
self.pca_data_transposed_aslist = self.pca_data_transposed.tolist()
for x, y, names in zip(self.pca_data_transposed_aslist[0],
self.pca_data_transposed_aslist[1], self.keys):
self.client.send_message('/data', [names, x, y])
printp('Starting PCA')
data = np.array(list(raw_data))
######### Dimensionality Reduction ##########
scaler = StandardScaler()
scaler.fit(data) # Fit Data with StandardScaler() (get mean and stddev)
# Transform original data set based on the .fita
data = scaler.transform(data)
pca = decomposition.PCA(n_components=2)
pca.fit(data)
data = pca.transform(data)
data_transposed = data.transpose() # X as one list, Y as another
data_transposed = np.ndarray.tolist(data_transposed)
# Normalisation
data = norm_np(data)
out_dict = {}
for key, value in zip(keys, data):
out_dict[key[:-5]] = list(value)
write_json(os.path.join(parent, 'sketches', 'livedb_pca.json'), out_dict)
# Write out pca data to stereo wav file
printp('Writing data to output files')
data = data.astype('float32')
wavfile.write(os.path.join(parent, 'sketches', 'livedb_pca.wav'), 44100,
data)
printp('Done')
def normalise_cols(self, slash_filter):
try:
max_post('Normalising data.', self.client)
self.data = norm_np(self.data)
except:
max_post('There is no data currently loaded.', self.client)