By on

A Gentle Introduction to Machine Learning for Product Managers

IMG_6783

(B) I taught a machine learning class to Product Managers two years ago. I have slightly updated the content of that class. A lot of Product Managers working on machine learning are contacting me on LinkedIn. So, I am hoping that the following materials might be useful to someone.

Class 1

ML University for PM Class 1

Videos:

Readings:

Homework:

  • 1. Select a few ML applications that you are using every day on your desktop or your mobile phone. Classify them into supervised or unsupervised learning. Try to identify the source of the data. Try to define how as a product/program manager or a product designer, you will consider those apps successful and which metrics, you might use to measure the app success. Please put your thoughts in a google document that you can share.
  • 2. Propose a few applications that could be developed using an estimation (based on a regression), a classification (based on a logistic regression), and a segmentation (based on a K-means clustering). Please put your thoughts in a google document that you can share.

Class 2

ML University for PM Class 2

Videos:

  • A machine learning primer: https://www.youtube.com/watch?v=1M09i0f3ruI&t=22s
  • (Please watch that video a couple of times, until you get it 🙂 We have not covered yet the K-nearest neighbor and Support Vector Machine algorithms but still try to understand those two algos)

Readings:

Optional (and advanced but interesting):

Homework:

  • 1. You have been hired to be the first Product Designer, Product Manager, or Program Manager for the Intelligence Platform at Sony Kamaji. Your first assignment is to lead the development of an ML application for classifying Sony games. To get started on that assignment, please propose a small draft of your proposed plan for that app. Please put your thoughts in the same google document that you use for your previous homework.
  •  2.  Since you completed ahead of schedule and with a lower budget your classification app for Sony games, you are now being asked to develop a segmentation app of SONY users. That app could leverage the labels of your classification apps. Please put your thoughts in the same google document that you use for your previous homework.
  • Note: You can be as creative as you want in terms of the product plan that you will use to drive the development of those apps but at least describe the “what” and “why” of the apps, the data sources/features, the possible models, and any KPIs. If you are running out of time, just do something simple in a few sentences for each homework in ½ page. If you have time, do as much as you can 🙂

Class 3

ML University for PM Class 3

Videos and readings:

Read and watch as much as you can 🙂

Google ML class:

Optional (for those with a math background):

Homework:

Answer the following questions. Note: please limit your answers to a few concise sentences:

  • What is a labeled training set?
  • What are the two most common tasks that a supervised machine learning model can do?
  • Give the name of a common task that an unsupervised machine learning model can do?
  • What is the difference between a model parameter and a learning algorithm hyperparameter?
  • If your model performs great on the training data but generalizes poorly to new instances, what is happening? Can you mention some possible solutions
  • What is a test data set and why would you want to use it?
  • Why feature engineering is a critical task for a data scientist to perform in order to develop a model that provides some good predictions?
  • When training a model, a data scientist might use an algorithm to reduce the cost or loss function of the model? What is in that case, the cost function?
  • What are the advantages of using a deep learning model compared to other traditional machine learning models?
  • Why deep learning models are somewhat considered as a “black box” model compared for instance to decision trees that are considered as a “white box” model?

Class 4

ML University for PM Class 4

Readings:

Deep learning:

Videos:

Training a neural network:

Homework:

Final project to be presented in Class 7

  • Each student should select an ML application either from her or his organization or outside her or his organization
  • The project deliverables shall include least the following sections:
    • ML application:
      • What, why, business case, business plan…
      • Use cases, user experiences, features, requirements, KPIs…
    • Models:
      • Model predictions
      • Machine learning tasks
      • Type of learning
      • Sucess metrics and KPIs
    • Features:
      • Feature sharing
    • Data sets
      • Data lakes
      • Data sources
      • Batch data
      • Streaming data
    • Machine learning pipeline:
      • Training
      • Inference
      • Experimentation
      • Feature store

Class 5

ML University for PM Class 5

Machine learning pipelines

Google:

Readings:

Videos:

Facebook:

Readings:

Uber:

Readings:

Videos:

Twitter:

Readings:

Airbnb:

Video:

Recommender systems:

Videos:

Readings:

  • A field study of video recommendations: newest, most similar, most relevant (easy reading):
  • Matrix factorization (not too difficult reading)
  • Factorization machines (difficult reading)

Homework:

Final project to be presented in Class 7

Class 6

ML University for PM Class 6

Videos:

Readings:

Homework:

Final project to be presented in Class 7

Class 7

Final project presentation:

  • Show time 🙂

 

Note: The picture above is a lemon from one of my citrus trees.

Copyright © 2005-2020 by Serge-Paul Carrasco. All rights reserved.
Contact Us: asvinsider at gmail dot com

Categories: Artificial Intelligence, Machine Learning, Product Management