NoelKennedy

import keras.backend as K

dtype='float16'
K.set_floatx(dtype)

# default is 1e-7 which is too small for float16.  Without adjusting the epsilon, we will get NaN predictions because of divide by zero problems
K.set_epsilon(1e-4) 

NoelKennedy

import math
import matplotlib
import matplotlib.pyplot as plt
from matplotlib.ticker import PercentFormatter

# x is a vector representing a sentence.  Sentence is an vector of integers.
# this method returns a matrix Z of size number_of_tokens_in_sentence * number_of_tokens_in_sentence 
# st that z[i][j] == x[j] except when i==j, z[i][j] == 0.  ie the token is blanked out with the padding indicator
def create_occlusion_batch(x):
    number_tokens_in_sentence=x.shape[0]

NoelKennedy

def make_random_input_sentence(sequence_length=300):
    import random
    input_words=list()
    input_sentence=list()
    for i in range(0,sequence_length):
        word=random.choice(list(sg.wv.vocab.keys()))
        input_words.append(word)
        vector=sg[word]
        input_sentence.append(vector)
    disease_token_bit=np.zeros((sequence_length,1))

NoelKennedy

np.random.normal(loc=0.0, scale=1.0)

NoelKennedy

def make_random_input_sentence(sequence_length=300,word2vec):
    import random
    input_words=list()
    input_sentence=list()
    for i in range(0,sequence_length):
        word=random.choice(list(sg.wv.vocab.keys()))
        input_words.append(word)
        vector=word2vec[word]
        input_sentence.append(vector)
    input_sentence=np.array(input_sentence)

NoelKennedy

# converts array of embedding intergers to a list of tokens that those integers represent
# the disease token is shown in upper case


# x_1 = array of embedding integers
# x_2 = array of disease phrase indicators
def embeddings_to_tokens(x_1,x_2,retain_null=True):
    tokens=list([embedding_to_token_map[embedding_id] for embedding_id in x_1])
    for i in range(0,x_2.shape[0]):
        if x_2[i]==1:

NoelKennedy

    public interface IQueryString {
        Option<string> GetValue(string key);
    }

NoelKennedy

public abstract partial class Either<T, U> {...}
//here are some business rules
public Either<LoanDeclinedReason, LoanRequest> CheckCustomerCredit(LoanRequest request) {
    if (request.Customer.CreditScore > 8) return Either.Right(request);
    return Either.Left(LoanDeclinedReason.InsufficientCreditScore);
}

public Either<LoanDeclinedReason, LoanRequest> CheckGoldPrerequisiteCustomer(LoanRequest request) {
    if (request.Value < 10000) return Either.Right(request); //dont need gold status for less than 10,000
    if (request.Customer.GoldCustomer) return Either.Right(request);

NoelKennedy

public static decimal GetTaxes(decimal salary)
{
    var taxBands = new[]
        {
            new Tuple<Decimal, Decimal, Decimal>(0, 5070, 0.1m),
            new Tuple<Decimal, Decimal, Decimal>(5070, 8660, 0.14m),
            new Tuple<Decimal, Decimal, Decimal>(8660, 14070, 0.23m),
            new Tuple<Decimal, Decimal, Decimal>(14070, 21240, 0.3m),
            new Tuple<Decimal, Decimal, Decimal>(21240, 40230, 0.33m),
            new Tuple<Decimal, Decimal, Decimal>(40230, Decimal.MaxValue, 0.45m)

NoelKennedy

class JsonpViewRenderer extends JsonViewRenderer {

  var nameOfCallbackParameter = "callback"

  private def jsonWithPadding(renderJson: => Unit, request: Request, response: Response) = {
    val callbackOption = request.getStringParameter(nameOfCallbackParameter)
    callbackOption match {
      case None => renderJson
      case Some(callback) => {
        response.getWriter.write(callback + "(")