this-is-richard’s gists

this-is-richard / plot_results.py

Created April 26, 2019 04:24

	def plot_results(history, with_acc=False):
	loss = history.history['loss']
	val_loss = history.history['val_loss']

	epochs = range(1, len(val_loss) + 1)

	plt.plot(epochs, loss, 'bo', label='Training loss')
	plt.plot(epochs, val_loss, 'b', label='Validation loss')
	plt.title('Training and validation loss')
	plt.legend()

this-is-richard / normalize_shuffle_data.py

Last active April 25, 2019 11:41

	from sklearn.preprocessing import StandardScaler

	training_sameples = int(total_samples // 2)
	scaler = StandardScaler()

	unused_col = ['Id']

	scaler.fit(df.drop(columns=unused_col).iloc[:training_sameples])
	normalized_df = scaler.transform(df.drop(columns=unused_col))

this-is-richard / pd_get_dummies.py

Created April 25, 2019 09:57

	for col in categorical:
	df[col] = df[col].map(lambda x: str(x))

	df = pd.get_dummies(df, drop_first=True)

this-is-richard / null_counts.py

Last active April 26, 2019 04:09

	def null_counts(df):
	null_counts = df.isnull().sum(axis = 0) / len(df)
	null_counts.plot.bar(figsize=(15, 4))
	print(null_counts.sort_values(ascending=False).head(10))

this-is-richard / stacked_rnns_binary_classification.py

Created April 25, 2019 05:40

	model = models.Sequential()
	model.add(layers.LSTM(32, return_sequences=True,
	input_shape=(num_timesteps, num_features)))
	model.add(layers.LSTM(32, return_sequences=True))
	model.add(layers.LSTM(32))
	model.add(layers.Dense(num_classes, activation='sigmoid'))

	model.compile(optimizer='rmsprop', loss='binary_crossentropy')

this-is-richard / single_rnn_binary_classification.py

Created April 25, 2019 05:39

	model = models.Sequential()
	model.add(layers.LSTM(32, input_shape=(num_timesteps, num_features)))
	model.add(layers.Dense(num_classes, activation='sigmoid'))

	model.compile(optimizer='rmsprop', loss='binary_crossentropy')

this-is-richard / image_classification.py

Created April 25, 2019 05:39

	model = models.Sequential()
	model.add(layers.SeparableConv2D(32, 3, activation='relu',
	input_shape=(height, width, channels)))
	model.add(layers.SeparableConv2D(64, 3, activation='relu'))
	model.add(layers.MaxPooling2D(2))

	model.add(layers.SeparableConv2D(64, 3, activation='relu'))
	model.add(layers.SeparableConv2D(128, 3, activation='relu'))
	model.add(layers.MaxPooling2D(2))

this-is-richard / regression.py

Created April 25, 2019 05:38

	model = models.Sequential()
	model.add(layers.Dense(32, activation='relu', input_shape=(num_input_features,)))
	model.add(layers.Dense(32, activation='relu'))
	model.add(layers.Dense(num_values))

	model.compile(optimizer='rmsprop', loss='mse')

this-is-richard / multi_label_categorical_classification.py

Created April 25, 2019 05:37

	model = models.Sequential()
	model.add(layers.Dense(32, activation='relu', input_shape=(num_input_features,)))
	model.add(layers.Dense(32, activation='relu'))
	model.add(layers.Dense(num_classes, activation='sigmoid'))

	model.compile(optimizer='rmsprop', loss='binary_crossentropy')
	# targets should be k-hot encoded

this-is-richard / single_label_categorical_classification.py

Created April 25, 2019 05:36

	model = models.Sequential()
	model.add(layers.Dense(32, activation='relu', input_shape=(num_input_features,)))
	model.add(layers.Dense(32, activation='relu'))
	model.add(layers.Dense(num_classes, activation='softmax'))

	model.compile(optimizer='rmsprop', loss='categorical_crossentropy')

Richard Ng this-is-richard