Feb 14, 2018 · fast forward food labs

Probabilistic Cookies!

In the spirit of Valentine’s Day, we at Fast Forward Labs thought it would be fun to bake cookies for our sweethearts. Being DIY nerds, we thought we’d math it up a bit.

We used python to generate probability distributions and matplotlib to check our distributions. Then we wrote a python function to generate a SCAD file defining three-dimensional shapes from the distributions. Using OpenSCAD, an open-source CAD program, we checked the 3D models and exported them to STL files for printing. We used a 5th Generation MakerBot Replicator to print our 3D models. And we baked cookies. Here’s one of our office dogs (and my best friend), Dogface, admiring the results.

There were a number of challenges involved in generating 3D models and printing them. Here’s how the basic process went.

We chose a beta distribution for our first prototype because it’s well behaved for purposes of making a 3D printed object.

  • A beta distribution only has values from 0 to 1. This gives us a fixed-width shape to work with. (Compare a Gaussian distribution, which has long tails on both sides and thus may not normalize to a good shape across 0 to 1.)
  • The area under the curve of a beta distribution is necessarily 1, which helps keep the shape from getting too eccentric while allowing flexibility in choice of parameters, and thus a wider range of shapes.
  • If you take two beta distributions, put them x-axis to x-axis, and squint, they look a bit like a heart.

We created a beta distribution in python.

import numpy as np
from scipy.stats import beta
import matplotlib.pyplot as plt

# choose enough points to have a relatively smooth curve without 
# creating so many facets that the 3D printer is slow
# choose an odd number so there is a definitive peak to the curve
numpoints = 35

# set up linspace
X = np.linspace(0, 1, numpoints)

# beta distribution parameters
a, b = 2, 1.6

# get beta distribution array
betadist = beta.pdf(X, a, b)

And plotted that distribution with some suitable parameters.

# turn off the axes so we only see the curve

# plot the curve itself
plt.plot(X, betadist)

Once we found a set of parameters that we liked, we used the points in the distribution in python to create an OpenSCAD-formatted SCAD file. A SCAD file is a readable text file that contains a combination of points in space that define shapes and instructions to manipulate those shapes. We made two 2D copies of the distribution, sized one up a bit, and centered them together. We extruded them both into 3D, one linear, and the other with a slight “cone” projection. This gave us one relatively thin edge for piercing cookie dough. Then we subtracted one shape from the other to make a hollow in the larger shape. Rendered in 3D in OpenSCAD, it looks like this:

Note that the 3D model looks more eccentric than the python-generated plot. This is because the plot had axes of differing scales.

With a 3D model in hand, it was time to print a physical object. We exported our SCAD object into a stereolithography (STL) file, which we then imported into MakerBot’s printing software.

We made a final print file and extruded our first prototypes at NYC Resistor our friendly neighborhood hackerspace. 3D printing can take a while for large shapes, so we started out small.

Here are the first results (we had begun tinkering with a Gaussian distribution at that point):

Of course these first test shapes are too small for cookies. After a lot of tinkering and refinement, we ended up with beta, Gaussian, and Poisson distribution shapes scaled up for cookie size (about 100mm high).

We made some cookie dough and got to business.

Note that the Poisson distribution (printed in white) has big solid areas at the top. This makes it more of a cookie-dough perturber than a cookie cutter. Those solid areas are an artifact of my own ignorance of OpenSCAD. The “clever” OpenSCAD scale() approach I had been using was a hack; I later learned that the offset() function is the correct solution.

OpenSCAD issues aside, the cookies turned out fine. Here are some of the results, some decorated with axes and histograms.

If you feel inclined to give it a shot yourself, here’s a jupyter notebook with some code to get you started.

Happy Valentine’s Day from Fast Forward Labs!

Read more

Feb 28, 2018 · newsletter
Jan 26, 2018 · newsletter

Latest posts

Nov 15, 2020 · post

Representation Learning 101 for Software Engineers

by Victor Dibia · Figure 1: Overview of representation learning methods. TLDR; Good representations of data (e.g., text, images) are critical for solving many tasks (e.g., search or recommendations). Deep representation learning yields state of the art results when used to create these representations. In this article, we review methods for representation learning and walk through an example using pretrained models. Introduction Deep Neural Networks (DNNs) have become a particularly useful tool in building intelligent systems that simplify cognitive tasks for users. more
Jun 22, 2020 · post

How to Explain HuggingFace BERT for Question Answering NLP Models with TF 2.0

by Victor · Given a question and a passage, the task of Question Answering (QA) focuses on identifying the exact span within the passage that answers the question. Figure 1: In this sample, a BERTbase model gets the answer correct (Achaemenid Persia). Model gradients show that the token “subordinate ..” is impactful in the selection of an answer to the question “Macedonia was under the rule of which country?". This makes sense .. good for BERTbase. more
Jun 16, 2020 · notebook

Evaluating QA: Metrics, Predictions, and the Null Response →

by Melanie · A deep dive into computing QA predictions and when to tell BERT to zip it! In our last post, Building a QA System with BERT on Wikipedia, we used the HuggingFace framework to train BERT on the SQuAD2.0 dataset and built a simple QA system on top of the Wikipedia search engine. This time, we’ll look at how to assess the quality of a BERT-like model for Question Answering.
May 19, 2020 · notebook

Building a QA System with BERT on Wikipedia →

by Melanie · So you’ve decided to build a QA system. You want to start with something simple and general so you plan to make it open domain using Wikipedia as a corpus for answering questions. You want to use the best NLP that your compute resources allow (you’re lucky enough to have access to a GPU) so you’re going to focus on the big, flashy Transformer models that are all the rage these days.
Apr 28, 2020 · notebook

Intro to Automated Question Answering →

by Melanie · Welcome to the first edition of the Cloudera Fast Forward blog on Natural Language Processing for Question Answering! Throughout this series, we’ll build a Question Answering (QA) system with off-the-shelf algorithms and libraries and blog about our process and what we find along the way. We hope to wind up with a beginning-to-end documentary that provides:
Apr 1, 2020 · newsletter

Enterprise Grade ML

by Shioulin · At Cloudera Fast Forward, one of the mechanisms we use to tightly couple machine learning research with application is through application development projects for both internal and external clients. The problems we tackle in these projects are wide ranging and cut across various industries; the end goal is a production system that translates data into business impact. What is Enterprise Grade Machine Learning? Enterprise grade ML, a term mentioned in a paper put forth by Microsoft, refers to ML applications where there is a high level of scrutiny for data handling, model fairness, user privacy, and debuggability. more

Popular posts

Oct 30, 2019 · newsletter
Exciting Applications of Graph Neural Networks
Nov 14, 2018 · post
Federated learning: distributed machine learning with data locality and privacy
Apr 10, 2018 · post
PyTorch for Recommenders 101
Oct 4, 2017 · post
First Look: Using Three.js for 2D Data Visualization
Aug 22, 2016 · whitepaper
Under the Hood of the Variational Autoencoder (in Prose and Code)
Feb 24, 2016 · post
"Hello world" in Keras (or, Scikit-learn versus Keras)


In-depth guides to specific machine learning capabilities


Machine learning prototypes and interactive notebooks


A usable library for question answering on large datasets.

Explain BERT for Question Answering Models

Tensorflow 2.0 notebook to explain and visualize a HuggingFace BERT for Question Answering model.

NLP for Question Answering

Ongoing posts and code documenting the process of building a question answering model.

Interpretability Revisited: SHAP and LIME

Explore how to use LIME and SHAP for interpretability.


Cloudera Fast Forward is an applied machine learning reseach group.
Cloudera   Blog   Twitter