lakshmanok

Challenge	Automation solution	Keep It Simple Solution
Training-Serving Skew, where differences crop up between training and serving stacks	Feature Store	Transform Pattern
Clients are interested in a capability that involves multiple steps, only one of which is the ML model	ML Pipelines	Stateless Serving Function
The incoming data changes characteristics	Continuous evaluation	Scheduled evaluation
The correct answer changes over time	Continuous traini

lakshmanok

Challenge	Automation solution
Training-Serving Skew, where differences crop up between training and serving stacks	Feature Store
Clients are interested in a capability that involves multiple steps, only one of which is the ML model	ML Pipelines
The incoming data changes characteristics	Continuous evaluation
The correct answer changes over time	Continuous training
Model fairness issues arise	Mumbo Jumbo
Parts of the system aren’t changed when new code

lakshmanok

Challenge	Automation solution	Keep It Simple Solution	Complexity needed if …
Training-Serving Skew, where differences crop up between training and serving stacks	Feature Store	Transform Pattern	Features injected server-side
Clients are interested in a capability that involves multiple steps, only one of which is the ML model	ML Pipelines	Stateless Serving Function	Cascade of ML models
The incoming data changes characteristics	Continuous evaluation	Scheduled evaluat

lakshmanok

import pandas as pd
import numpy as np
import matplotlib.pyplot as plt
from matplotlib.backends.backend_pdf import PdfPages

def _draw_as_table(df, pagesize):
    alternating_colors = [['white'] * len(df.columns), ['lightgray'] * len(df.columns)] * len(df)
    alternating_colors = alternating_colors[:len(df)]
    fig, ax = plt.subplots(figsize=pagesize)
    ax.axis('tight')

lakshmanok

# Copyright 2021 Google LLC
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
#     https://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,

lakshmanok

# Copyright 2021 Google LLC
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
#     https://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,

lakshmanok

CREATE OR REPLACE TABLE publicdata.us_tracts AS
WITH populated_areas AS (
  SELECT 
  bounds, population_density
  FROM `ai-analytics-solutions`.publicdata.popdensity_sedac_rectangles
  WHERE 
  year = 2020 AND
  (tile = 'gpw_v4_population_density_rev11_2020_30_sec_1.asc' OR
   tile = 'gpw_v4_population_density_rev11_2020_30_sec_2.asc'
  ) AND

lakshmanok

PROJECT='ai-analytics-solutions'
BUCKET='ai-analytics-solutions-kfpdemo'
REGION='us-central1'

from datetime import datetime
import apache_beam as beam

def parse_nlp_result(response):
    return [
        # response, # entire string

lakshmanok

def create_geo(LATRES, LONRES, ORIGIN_LAT, ORIGIN_LON, rowno, startcol, endcol):
  # represent each rectangle by a polygon of its corners
  top = ORIGIN_LAT - rowno * LATRES
  bot = ORIGIN_LAT - (rowno+1) * LATRES
  left = ORIGIN_LON + startcol * LONRES
  right = ORIGIN_LON + endcol * LONRES

  poly = json.dumps({

lakshmanok

def create_rectangle_geo(LATRES, LONRES, ORIGIN_LAT, ORIGIN_LON, rowno, startcol, linedata):
  # find a rectangle of pixels with the same value, and represent that pixel instead
  endcol = startcol
  while (endcol < len(linedata)) and (linedata[endcol] == linedata[startcol]):
    endcol = endcol + 1
  return create_geo(LATRES, LONRES, ORIGIN_LAT, ORIGIN_LON, rowno, startcol, endcol)