hedgefair

# Author: HJ van Veen <[email protected]>
# Description: Experiment to learn a tSNE transformer for new 
#              test data with a multi-output GBM
#
# Idea first seen at lvdmaaten.github.io/tsne
# > [...] it is not possible to embed test points in an existing 
# > map [...]
# > A potential approach to deal with this would be to train 
# > a multivariate regressor to predict the map location from 
# > the input data.

hedgefair

from __future__ import print_function

import os

import numpy as np
from keras.layers import RepeatVector
from keras.layers.core import Dropout
from keras.layers.recurrent import LSTM
from keras.models import Sequential
from keras.models import load_model

hedgefair

A
ACO
AA
ACOM
AAN
ACOR
AAON
ACPW
AAP
ACTG

hedgefair

import keras.backend as K
from keras.engine import Layer
import tensorflow as tf


class SelU(Layer):

    def __init__(self, alpha=1.6732632423543772848170429916717, scale=1.0507009873554804934193349852946, **kwargs):
        super(SelU, self).__init__(**kwargs)
        self.alpha = K.cast_to_floatx(alpha)

hedgefair

A quick and dirty syntax translation / conversion reference guide to ease the transition between Python and Julia. This is not meant as a reference to the language. For that you should read the manual.

Some important differences

• Arrays in Julia are indexed starting from 1.
• In Julia classes (i.e. types) don't own methods. Methods are implementations of generic functions and are invoked in a "static style", i.e. instead of Python's str1.rstrip(), we will have rstrip( str1 ), instead of file1.close(), close( file1 ).

Some important similarities.

hedgefair

What

Roll your own iPython Notebook server with Amazon Web Services (EC2) using their Free Tier.

What are we using? What do you need?

• An active AWS account. First time sign-ups are eligible for the free tier for a year
• One Micro Tier EC2 Instance
• With AWS we will use the stock Ubuntu Server AMI and customize it.
• Anaconda for Python.
• Coffee/Beer/Time

hedgefair

''' Script for Cartpole using policy gradient via Chainer, two layer MLP, dropout, and rejection sampling of historical memories '''

import gym 

import numpy as np

import chainer
from chainer import optimizers
from chainer import ChainList, Variable
import chainer.functions as F

hedgefair

from keras.layers import GRU, initializations, K
from collections import OrderedDict

class GRULN(GRU):
    '''Gated Recurrent Unit with Layer Normalization
    Current impelemtation only works with consume_less = 'gpu' which is already
    set.

    # Arguments
        output_dim: dimension of the internal projections and the final output.

hedgefair

""" Trains an agent with (stochastic) Policy Gradients on Pong. Uses OpenAI Gym. """
import numpy as np
import cPickle as pickle
import gym

# hyperparameters
H = 200 # number of hidden layer neurons
batch_size = 10 # every how many episodes to do a param update?
learning_rate = 1e-4
gamma = 0.99 # discount factor for reward