1609010057-adaptation-rate.yaml

default_parameters:

  experiment: adaptive combination

  dataset_filename: enron-kayla
  label_type:
    ham_label: -1
    spam_label: 1

  classifier_fast:
    type: adaline
    training_parameters:
      learning_rate: 0.005
      initial_weights: null
    testing_parameters: {}

  classifier_slow:
    type: adaline
    training_parameters:
      learning_rate: 0.001
      initial_weights: null
    testing_parameters: {}

  training_parameters:
    window_size: 1
    window_operator: mean

  attack:
    parameters:
      percentage_samples_poisoned: .50
      start: .65
      duration: .20


parameter_ranges:

  - name: classifier fast
    key: [classifier_fast, type]
    values:
      - adaline
      - logistic regression

  - name: classifier slow
    key: [classifier_slow, type]
    values:
      - adaline
      - logistic regression

  - name: η fast
    key: [classifier_fast, training_parameters, learning_rate]
    values:
      - .01
      - .005

  - name: η slow
    key: [classifier_slow, training_parameters, learning_rate]
    values:
      - .001
      - .0005

  - name: attack
    key: [attack, type]
    values:
      - dictionary
      - empty

  - name: adaptation rate
    key: [training_parameters, adaptation_rate]
    values:
      - 0.5
      - 2

1609010058-adaptation-rate.yaml

default_parameters:

  experiment: adaptive combination

  dataset_filename: enron-kayla
  label_type:
    ham_label: -1
    spam_label: 1

  classifier_fast:
    type: adaline
    training_parameters:
      learning_rate: 0.005
      initial_weights: null
    testing_parameters: {}

  classifier_slow:
    type: adaline
    training_parameters:
      learning_rate: 0.001
      initial_weights: null
    testing_parameters: {}

  training_parameters:
    window_size: 1
    window_operator: mean

  attack:
    parameters:
      percentage_samples_poisoned: .50
      start: .65
      duration: .20


parameter_ranges:

  - name: classifier fast
    key: [classifier_fast, type]
    values:
      - adaline
      - logistic regression

  - name: classifier slow
    key: [classifier_slow, type]
    values:
      - adaline
      - logistic regression

  - name: η fast
    key: [classifier_fast, training_parameters, learning_rate]
    values:
      - .01
      - .005

  - name: η slow
    key: [classifier_slow, training_parameters, learning_rate]
    values:
      - .001
      - .0005

  - name: attack
    key: [attack, type]
    values:
      - dictionary
      - empty

  - name: adaptation rate
    key: [training_parameters, adaptation_rate]
    values:
      - 0.5
      - 2

1609010058.ipynb