ADGEfficiency

"""
A feedforward neural network aka multilayer perceptron.
Single hidden layer with relu as the activation function 
after the input & hidden layers.

Linear activation function after the output layer.

Built using 100% pure numpy.

Adam Green - adgefficiency.com

ADGEfficiency

# -*- coding: utf-8 -*-
"""
Created on Wed Jan 25 10:12:34 2017

@author: Adam Green
adam.green@adgefficiency.com
adgefficiency.com
"""

import os

ADGEfficiency

" ENCODING
set encoding=utf8
scriptencoding utf-8

" VUNDLE SETUP
" git clone https://github.com/VundleVim/Vundle.vim.git ~/.vim/bundle/Vundle.vim
set nocompatible              " be iMproved, required
filetype off                  " required

" set the runtime path to include Vundle and initialize

ADGEfficiency

import numpy as np
import pandas as pd

data = np.concatenate([
 np.random.normal(5, 1, 10000),
 np.random.normal(-5, 1, 10000),
 np.array([-20, 20] * 1000)
])

pd.DataFrame(data).plot(kind='hist', legend=None, bins=100)

ADGEfficiency

#  this search isn't wide enough
useless_search = sklearn.model_selection.GridSearchCV(
    sklearn.ensemble.RandomForestRegressor(n_estimators=10), param_grid={'n_estimators': [10, 15, 20]
)
#  this search is more informative
useful_search = sklearn.model_selection.GridSearchCV(
    sklearn.ensemble.RandomForestRegressor(n_estimators=10), param_grid={'n_estimators': [10, 100, 1000]

ADGEfficiency

#  bad
Var=1

#  good
var = 1

# bad
def adder ( x =10 ,y= 5):
    return  x+y
    

ADGEfficiency

standardized = (data - np.mean(data)) / np.std(data)

ADGEfficiency

standardized = (data - np.mean(data)) / np.std(data)

ADGEfficiency

standardized = (data - np.mean(data)) / np.std(data)

ADGEfficiency

normalized = (data - np.min(data)) / (np.max(data) - np.min(data))