jotterbach

import seaborn as sns
import numpy as np
import numpy.random as rd
import matplotlib.pyplot as plt
from scipy.spatial import distance as dist
import scipy.linalg as la
import itertools as it

%matplotlib inline

jotterbach

(*
The Evolution Strategy Optimizer is a Hill-Climbing routine based on stochastic gradient
estimation (not to be confused with stochastic gradient descent optimization). The basic idea
is to perturb the current best parameter value using Gaussian noise and explore the function
values of the blackbox objective in the vicinity of the parameters. From these explorations one
can estimate the gradient and perform a gradient update step.

More information can be found in:
- T. Salimans et al. Evolution Strategies as a Scalable Alternative to Reinforcement Learning,
    arXiv:1703.03864, https://arxiv.org/pdf/1703.03864.pdf

jotterbach

(*
This is a simple solution to the Run-Length Encoding problem described
in Watrophy #24 (http://www.watrophy.com/posts/24-Run-Length-Encoding.html)
*)

open Core

let rle int_list =
  let increment tup n =
    match tup with

jotterbach

(** The fundamental idea behind this sort algorithm is to iteratively remove nodes from a DAG
 * who do not possess ancestors and hence can be considered roots. We add them to an ordered list
 * and get a topological sort of the dependency graph. *)

open Core
open Graph.Pack.Digraph


(** Simple helper to create a OCamlGraph with directionality *)
let create_digraph nodes edges =

jotterbach

open Graph
open Core
open Owl

module G = Persistent.Graph.Concrete(
  struct
    type t = int
    let equal a b = phys_equal a b
    let compare a b = Pervasives.compare a b
    let hash a = a

jotterbach

from torch.autograd.gradcheck import zero_gradients


def compute_jacobian(inputs, output):
    """
    :param inputs: Batch X Size (e.g. Depth X Width X Height)
    :param output: Batch X Classes
    :return: jacobian: Batch X Classes X Size
    """
    assert inputs.requires_grad

jotterbach

def make_grid(arr: np.ndarray, nrows: int, pad_size: int = 2, pad_value: int = 0):
    """
    Expected array layout is (N, C, H, W)
    """
    arr = np.pad(arr,
                 ((0,0),
                  (0,0),
                  (pad_size // 2, pad_size // 2),
                  (pad_size // 2, pad_size // 2)),
                 constant_values=(pad_value,))

jotterbach

import numpy as np


def conv_out_shape(in_shape: tuple, kernel_size:int, dilation:int, stride:int, padding:int):
    """
    Convenience function for computing the output shape of an image tensor, when applying symmetric kernels, strides, etc.
    """
    out_dim = []
    for d in in_shape:
        _d_out = d + 2 * padding - dilation * (kernel_size - 1) - 1

jotterbach

#!/bin/bash

### steps ####
# verify the system has a cuda-capable gpu
# download and install the nvidia cuda toolkit and cudnn
# setup environmental variables
# verify the installation
###

### to verify your gpu is cuda enable check