timvieira

timvieira

timvieira

# -*- coding: utf-8 -*-
"""
Backprop as the method of Lagrange multiplers (and even the implicit function
theorem).
"""
from __future__ import division
import numpy as np
from arsenal.alphabet import Alphabet
from arsenal.math.checkgrad import finite_difference

timvieira

"""
Simple example of manually performing "automatic" differentiation
"""
import numpy as np
from numpy import exp, sin, cos


def f(x, with_grad=False):
    # Need to cache intermediates from forward pass (might not use all of them).
    a = exp(x)

timvieira

# Efficient passive aggressive updates for multi-class classification
#
# Original article:
# "Column squishing for multiclass updates"
# https://nlpers.blogspot.com/2017/08/column-squishing-for-multiclass-updates.html


from __future__ import division
import numpy as np
import scipy.optimize

timvieira

"""
Memory-efficient backpropagation in an RNN.

Accompanies blog post:
http://timvieira.github.io/blog/post/2016/10/01/reversing-a-sequence-with-sublinear-space/
"""

import numpy as np
from arsenal.math.checkgrad import fdcheck

timvieira

"""
Memory-efficient backpropagation in an RNN.

Accompanies blog post:
http://timvieira.github.io/blog/post/2016/10/01/reversing-a-sequence-with-sublinear-space/
"""

import numpy as np
from arsenal.math.checkgrad import fdcheck

timvieira

"""
Cartoon version of Jiawei's optimization problem.

Created [2017-02-17 Fri]

"""
import numpy as np
from scipy.optimize import fmin_bfgs

import autograd

timvieira

timvieira

import numpy as np
from numpy.random import uniform


def update(S, k, v):
    "Update value position `k` in time O(log n)."
    d = S.shape[0]
    i = d//2 + k
    S[i] = v
    while i > 0: