Zichen Wang wangz10

🎯

Focusing

Machine Learning Scientist

wangz10 / time_jax_functions.py

Created September 3, 2019 01:22

	In [1]: import numpy as np
	...: import jax.numpy as jnp
	...: from jax import random, jit

	In [2]: def slow_f(x):
	...: # Element-wise ops see a large benefit from fusion
	...: return x * x + x * 2.0
	...:
	...: # use XLA to compile the function
	...: fast_f = jit(slow_f)

wangz10 / pca_svd_relationship_np.py

Created March 16, 2019 14:34

	# Compute covariance matrix
	C = np.dot(X.T, X) / (n-1)
	# Eigen decomposition
	eigen_vals, eigen_vecs = np.linalg.eig(C)
	# SVD
	U, Sigma, Vh = np.linalg.svd(X,
	full_matrices=False,
	compute_uv=True)
	# Relationship between singular values and eigen values:
	print(np.allclose(np.square(Sigma) / (n - 1), eigen_vals)) # True

wangz10 / svd_np.py

Created March 16, 2019 14:29

Perform SVD with numpy

	def svd(X):
	# Data matrix X, X doesn't need to be 0-centered
	n, m = X.shape
	# Compute full SVD
	U, Sigma, Vh = np.linalg.svd(X,
	full_matrices=False, # It's not necessary to compute the full matrix of U or V
	compute_uv=True)
	# Transform X with SVD components
	X_svd = np.dot(U, np.diag(Sigma))
	return X_svd

wangz10 / pca_eigen_docomposition_np.py

Created March 16, 2019 14:25

Perform PCA by eigen decomposition

	def pca(X):
	# Data matrix X, assumes 0-centered
	n, m = X.shape
	assert np.allclose(X.mean(axis=0), np.zeros(m))
	# Compute covariance matrix
	C = np.dot(X.T, X) / (n-1)
	# Eigen decomposition
	eigen_vals, eigen_vecs = np.linalg.eig(C)
	# Project X onto PC space
	X_pca = np.dot(X, eigen_vecs)

wangz10 / interpolation_with_generative_model.py

Last active January 14, 2019 02:51

	def interpolate_from_a_to_b_for_c(model, X, labels, a=None, b=None, x_c=None, alpha=0.):
	'''Perform interpolation between two classes a and b for any sample x_c.
	model: a trained generative model
	X: data in the original space with shape: (n_samples, n_features)
	labels: array of class labels (n_samples, )
	a, b: class labels a and b
	x_c: input sample to manipulate (1, n_features)
	alpha: scalar for the magnitude and direction of the interpolation
	'''
	# Encode samples to the latent space

wangz10 / jointplot_w_hue.py

Last active September 26, 2018 16:18 — forked from ruxi/jointplot_w_hue.py

jointplot_w_hue

	__author__ = "lewis.r.liu@gmail.com"
	__copyright__ = "Copyright 2018, github.com/ruxi"
	__license__ = "MIT"
	__version__ = 0.0.1

	# update: Mar 5 , 2018
	# created: Feb 19, 2018
	# desc: seaborn jointplot with 'hue'
	# prepared for issue: https://github.com/mwaskom/seaborn/issues/365
	"""

wangz10 / .gitignore

Last active August 29, 2015 14:24 — forked from cornhundred/.gitignore

d3_clustergram