iskandr

import keras.backend as K
from keras.layers import Dense, Activation, Multiply, Add, Lambda
import keras.initializers

def highway_layers(value, n_layers, activation="tanh", gate_bias=-3):
    dim = K.int_shape(value)[-1]
    gate_bias_initializer = keras.initializers.Constant(gate_bias)
    for i in range(n_layers):     
        gate = Dense(units=dim, bias_initializer=gate_bias_initializer)(value)
        gate = Activation("sigmoid")(gate)

iskandr

import time
import numpy as np 
import sklearn.decomposition
import seaborn

RANK = 50
N_COLS = 1000

def evaluate_svd(svd_fn, reconstruct_fn, min_rows=100, max_rows=5000, n_samples=100, n_cols=N_COLS, rank=RANK, random_seed=0):
    np.random.seed(random_seed)

iskandr

  object ForLoop {
    def main(args: Array[String]) {
      var acc = 0L
      for (i <- 0L to 1000000000L) {
        acc += i
      }
      println("For loop!")
    }
  }

iskandr

iskandr

import numpy as np 

LeftCounts = np.array([150,150])
RightCounts = np.array([200,200])

NLeft = sum(LeftCounts)
NRight = sum(RightCounts)
NTotal = float(NLeft + NRight)
LL = np.dot(LeftCounts,LeftCounts)
RR = np.dot(RightCounts,RightCounts)

iskandr

def dilate_naive(x, k):
  m,n = x.shape
  y = np.empty_like(x)
  for i in xrange(m):
    for j in xrange(n):
      currmax = x[i,j]
      for ii in xrange(max(0, i-k/2), min(m, i+k/2+1)):
        for jj in xrange(max(0, j-k/2), min(n, j+k/2+1)):
          elt = x[ii,jj]
          if elt > currmax:

iskandr

def dilate_two_pass(x, k):
  m,n = x.shape
  y = np.empty_like(x)
  for i in xrange(m):
    for j in xrange(n):
      left_idx = max(0, i-k/2)
      right_idx = min(m, i+k/2+1) 
      currmax = x[left_idx, j]
      for ii in xrange(left_idx+1, right_idx):
        elt = x[ii, j]

iskandr

def dilate_two_pass(x, window_size):
  m,n = x.shape
  k,l = window_size 
  y = np.empty_like(x)
  for i in xrange(m):
    for j in xrange(n):
      left_idx = max(0, i-k/2)
      right_idx = min(m, i+k/2+1) 
      currmax = x[left_idx, j]
      for ii in xrange(left_idx+1, right_idx):

iskandr

def dilate_naive(x, (k,l)):
  m,n = x.shape
  y = np.empty_like(x)
  for i in xrange(m):
    for j in xrange(n):
      currmax = x[i,j]
      for ii in xrange(max(0, i-k/2), min(m, i+k/2+1)):
        for jj in xrange(max(0, j-l/2), min(n, j+l/2+1)):
          elt = x[ii,jj]
          if elt > currmax:

iskandr

def numpy_count_thresh(values, thresh):
  return np.sum(values < thresh)