charleslparker

import os
import sys

URL = "https://storage.googleapis.com/mledu-datasets/cats_and_dogs_filtered.zip"


def run_gpu_test(use_cuda):
    os.environ["TF_CPP_MIN_LOG_LEVEL"] = "3"

    if not use_cuda:

charleslparker

import java.io.File;
import java.io.FileInputStream;
import java.io.FileOutputStream;
import java.io.IOException;
import java.io.ObjectInputStream;
import java.io.ObjectOutputStream;
import java.io.Serializable;

import org.junit.Test;

charleslparker

Training dataset: https://bigml.com/shared/dataset/ymIL3HLZnUZCOhVCVEOlVNIpmQE
Test dataset: https://bigml.com/shared/dataset/rx2sKQFQgGuYbbtPfeI309xizBx

charleslparker

{
  "name": "Custom feature analyzer",
  "description": "Find the best features for modeling using a greedy algorithm",
  "kind": "script",
  "source_code": "analyze-features.whizzml",
  "inputs": [
    {
      "name": "dataset-id",
      "type": "dataset-id",

charleslparker

;; Get feature names given ids
(define (feature-names dataset-id ids)
  (let (fields (get (fetch dataset-id) "fields"))
    (map (lambda (id) (get-in fields [id "name"])) ids)))

(define (create-k-folds dataset-id k-folds)
  (let (k-fold-fn (lambda (x) (create-dataset
                               {"origin_dataset" dataset-id
                                "row_offset" x
                                "row_step" k-folds

charleslparker

;; This is a vanilla implementation of gradient boosting.  The main
;; function is at the bottom of the script, where it explains the
;; algorithm in some detail.

;; A constant added to the generated field names to let us know that
;; we generated them
(define boost-id "__bmlboost")

;; The names of the fields contain ground truth - if there are k
;; classes, this is k coluns, one for each class.  If the true class

charleslparker

Services and Algorithms

This section describes each of the services compared in the throwdown and the algorithms/models used.

BigML

Decision trees, both single and bagged.

charleslparker

#!/bin/bash

# Set credentials for BigML and for the US Census
BIGML_USERNAME=bigml
BIGML_API_KEY=****
CENSUS_API_KEY=****
BIGML_AUTH="username=$BIGML_USERNAME;api_key=$BIGML_API_KEY"

# Download demographic data into files.  Have to do it this way
# because the census API doesn't seem to allow returns of more