geffy

import numpy as np
%pylab inline


def generate_raw_weights(n, a, b, p):
    res = np.zeros((n, n))
    for i in range(n):
        for j in range(n):
            sq_dist = ((i - a)**2 + (j - b)**2) *1.0/ n**2
            res[i,j] = exp(-sq_dist*p)

geffy

def hard_normalizing(X):
    return (X - 0.5) / 0.5
  
    
def init_model():
    model = Sequential()

    model.add(Convolution2D(64, 3, 3, border_mode='valid', input_shape=(3, 64, 64)))
    model.add(Activation('relu'))
    model.add(MaxPooling2D(pool_size=(2, 2)))

geffy

import gym
import numpy as np

env = gym.make('FrozenLake8x8-v0')
env.reset()

# find terminals
def find_terminals(mdp_raw):
    terminals = set()
    for src_state, node in mdp_raw.items():

geffy

import gym
import numpy as np

env = gym.make('FrozenLake8x8-v0')
env.reset()

# find terminals
def find_terminals(mdp_raw):
    terminals = set()
    for src_state, node in mdp_raw.items():

geffy

# Solving as MDP using Value Iteration Algorithm

import gym
import numpy as np

def iterate_value_function(v_inp, gamma, env):
    ret = np.zeros(env.nS)
    for sid in range(env.nS):
        temp_v = np.zeros(env.nA)
        for action in range(env.nA):

geffy

import gym
import numpy as np

env = gym.make('FrozenLake8x8-v0')
env.reset()

# policy obtained from solving MDP
policy = np.array(
    [3, 2, 2, 2, 2, 2, 2, 2,
     3, 3, 3, 3, 3, 3, 3, 2,

geffy

import numpy as np
import tensorflow as tf
import os
import glob

def tf2npz(tf_path, export_folder='/ssd/yt8m/data_npz/'):
    vid_ids = []
    labels = []
    mean_rgb = []
    mean_audio = []

geffy

geffy

# -*- coding: utf-8 -*-
"""
Created on Mon Sep 23 23:16:44 2017

@author: Marios Michailidis

This is an example of a simple method that performs bagging

"""

geffy

# -*- coding: utf-8 -*-
"""
Created on Mon Sep 23 23:16:44 2017

@author: Marios Michailidis

This is an example that performs stacking to improve mean squared error
This examples uses 2 bases learners (a linear regression and a random forest)
and  linear regression (again) as a meta learner to achieve the best score.
The initial train data are split in 2 halves to commence the stacking.