jaor

(ns histogram
  (:use
    [clojure.contrib.math :only [abs]]
    [clojure.contrib.priority-map]))

(defn- bin-weight [bin]
  (* (double (first bin)) (double (second bin))))

(defn- combine-bins [prev-bin next-bin]
  (let [first-weight (bin-weight prev-bin)

jaor

JSON schemas for models

• model-schema.json A generic ML model, containing fields shared by most models despite of their concrete type. It uses:
  • sample-schema.json The schema for dataset sampling specifications
  • field-collection-schema.json Auxiliary schema describing a collection of field (or "properties") descriptors
  • generic-field-schema.json Properties shared by all fields, regardless of their type.
  • field-schema.json The union schema of all field descriptor types, with their specific properties.
• tree-model-schema.json A specialization of the model schema to decision tree models. It uses:
• node-schema The schema for the nodes in a decision tree

jaor

(define (create-evaluations project-id)
  (let (src-ids (resource-ids (list-sources {"project" project-id})))
    (for (src-id src-ids)
      (let (ds-id (create-dataset src-id)
            [train-id test-id] (create-random-dataset-split ds-id 0.8)
            ens-id (create-ensemble train-id {"number_of_models" 100}))
        (create-evaluation test-id ens-id)))))

(define result (wait* (create-evaluations "project/xxxxxxxxxx")))

jaor

{
  "name": "Muti-model predictions",
  "kind": "script",
  "description": "Generate predictions for a fixed number of models",
  "source_code": "multibatch.whizzml",
  "inputs":[
    {
      "name": "model1",
      "type": "model-id",
      "description": "first model"

jaor

{
  "name": "Prune terms",
  "kind": "script",
  "description": "Removes from a text fields all entries with low coverage",
  "source_code": "script.whizzml",
  "inputs":[
    {
      "name": "dataset-id",
      "type": "dataset-id",
      "description": "The original dataset"

jaor

{
  "name": "Filtered timeseries",
  "kind": "script",
  "description": "Takes a filter field and an objective field and creates a timeseries for each category in the filter field, collecting all resulting forecasts in a new dataset",
  "source_code": "script.whizzml",
  "inputs":[
    {
      "name": "dataset",
      "type": "dataset-id",
      "description": "Input dataset"

jaor

{
  "name": "Name topics",
  "kind": "script",
  "description": "Give a name to all topics in a topicmodel",
  "source_code": "script.whizzml",
  "imports":[
    ],
  "inputs":[
    {
      "name": "topic-model",

jaor

(define (model-range dataset from to)
  (create-model dataset {"range" [from to]}))

(define (eval-range dataset model from to)
  (let (ev-id (create-evaluation dataset model {"range" [from to]}))
    [ev-id ((fetch (wait ev-id)) ["result" "model" "average_phi"])]))

(define (size-evaluations dataset-id steps)
  (let (ds (fetch dataset-id)
        rows (ds "rows")

jaor

{
  "name": "select-important",
  "kind": "script",
  "description": "Select the important features from an existig moddel to create a new one",
  "source_code": "script.whizzml",
  "imports":[
    ],
  "inputs":[
    {
      "name": "model-id",

jaor

{
  "name": "Mark non-preferred",
  "kind": "script",
  "description": "Given a dataset, mark as non-preferred fields starting with a prefix",
  "source_code": "script.whizzml",
  "imports":[
    ],
  "inputs":[
    {
      "name": "dataset-id",