• You Can’t Escape Hyperparameters and Latent Variables: Machine Learning as a Software Engineering Enterprise [Video & Slides. Talk starts at 28:55]
  In this talk, I will argue that the community needs to adopt systematic approaches for creating robust artifacts that contribute to larger systems that impact the real human world. I will share perspectives from multiple researchers in machine learning, theory, computer perception, and education; discuss with them approaches that might help us to develop more robust machine-learning systems; and explore scientifically interesting problems that result from moving beyond narrow machine-learning algorithms to complete machine-learning systems...

• Understanding the role of individual units in a deep neural network
  Deep neural networks excel at finding hierarchical representations that solve complex tasks over large datasets. How can we humans understand these learned representations? In this work, we present network dissection, an analytic framework to systematically identify the semantics of individual hidden units within image classification and image generation networks...

• From 0 to 60 (Models) in Two Years: Building Out a Data Science Function
  In 2018, I was hired to help WW (formerly Weight Watchers) build out a data science function. At the time, there was just a single junior data scientist and zero models in production. Today, we have a thriving, talented team of nine data scientists, more than 60 live models, and various data products in the hands of millions of WW members. When so many data science teams fail to reach their potential, what are some of the factors that led to our success?...

Data scientists are in demand on Vettery

Get discovered by one of the thousands of companies using Vettery to grow their data science teams. Here’s how it works: create a profile, name your salary, and connect with hiring managers from startups to Fortune 500 companies.
Get started - it’s completely free for job-seekers!

• Causal Curve: tools to perform causal inference using observational data when the treatment of interest is continuous
  There are many implemented methods to perform causal inference when your intervention of interest is binary, but few methods exist to handle continuous treatments...This is unfortunate because there are many scenarios (in industry and research) where these methods would be useful. For example, when you would like to: a) Estimate the causal response to increasing or decreasing the price of a product across a wide range, b) Understand how the number of minutes per week of aerobic exercise causes positive health outcomes, c) Estimate how decreasing order wait time will impact customer satisfaction, after controlling for confounding effects...etc...This library attempts to address this gap, providing tools to estimate causal curves (AKA causal dose-response curves)...

• Using JAX to accelerate our research
  Recently, we've found that an increasing number of projects are well served by JAX, a machine learning framework developed by Google Research teams. JAX resonates well with our engineering philosophy and has been widely adopted by our research community over the last year. Here we share our experience of working with JAX, outline why we find it useful for our AI research, and give an overview of the ecosystem we are building to support researchers everywhere...

• Topic Modeling the Mishnah
  In this project, I sought to implement a series of Natural Language Processing techniques to examine ancient Jewish religious texts (namely the Mishnah). I had two subjects that I wished to examine. The first was the use of topic-modeling techniques to match the already established categories in the Mishnah. This involved dimensionality reduction of a “bag of words” model of the corpus. The second was an examination of the intra-text agreement of opinions within the Mishnah. The relevance of this second goal will become more clear with further examination of the nature of the documents...

• Transformers for Image Recognition at Scale
  While convolutional neural networks (CNNs) have been used in computer vision since the 1980s, they were not at the forefront until 2012 when AlexNet surpassed the performance of contemporary state-of-the-art image recognition methods by a large margin...Looking forward to the next generation of scalable vision models, one might ask whether this domain-specific design is necessary, or if one could successfully leverage more domain agnostic and computationally efficient architectures to achieve state-of-the-art results...As a first step in this direction, we present the Vision Transformer (ViT), a vision model based as closely as possible on the Transformer architecture originally designed for text-based tasks...

• Scaling healthcare through AI
  Curai’s mission is to “provide the world’s best healthcare to everyone”. In order to do so, we strive to scale the reach of an individual physician by creating tools that automate large portions of a medical interaction under their supervision...To accomplish our goals while following the previous premises, we are using a combination of: a) State-of-the-art AI in the wild and b) AI with experts-in-the-loop...In the rest of this post, we will describe these two dimensions and the techniques that support them...

• The NetHack Learning Environment
  Progress in Reinforcement Learning (RL) algorithms goes hand-in-hand with the development of challenging environments that test the limits of current methods. While existing RL environments are either sufficiently complex or based on fast simulation, they are rarely both. Here, we present the NetHack Learning Environment (NLE), a scalable, procedurally generated, stochastic, rich, and challenging environment for RL research based on the popular single-player terminal-based roguelike game, NetHack...

• On the Modularity of Hypernetworks
  In the context of learning to map an input I to a function hI:X→ℝ, two alternative methods are compared: (i) an embedding-based method, which learns a fixed function in which I is encoded as a conditioning signal e(I) and the learned function takes the form hI(x)=q(x,e(I)), and (ii) hypernetworks, in which the weights θI of the function hI(x)=g(x;θI) are given by a hypernetwork f as θI=f(I). In this paper, we define the property of modularity as the ability to effectively learn a different function for each input instance I...

• Combining Dask and PyTorch for Better, Faster Transfer Learning
  Data parallelism within a single machine is a reasonably well-documented method for optimizing deep learning training performance, particularly in PyTorch. However, taking the step from one machine to training a single neural net on many machines at once can seem difficult and complicated...This tutorial will demonstrate first, that GPU cluster computing to conduct transfer learning allows the data scientist to significantly improve the effective learning of a model; and second, that implementing this in Python is not as hard or scary as it sounds, especially with our new library, dask-pytorch-ddp...

• Board Game Image Recognition using Neural Networks
  Utilizing computer vision techniques and convolutional neural networks (CNN), the algorithms created for this project classify chess pieces and identify their location on a chessboard. The final application saves images throughout to visualize the performance and outputs a 2D image of the chessboard to see the results (see below)...

Feature store: The data platform for building, deploying, and using ML features

The Uber Michelangelo team built the first feature store to scale Uber’s Machine Learning to 1000s of production models in just a few years. Feature stores have now become an essential part of the modern stack for operational ML. They bring DevOps principles to ML data, and allow data scientists to build great ML features, get them to production instantly, and share them across teams. Mike Del Balso, Co-Founder of Tecton, and Willem Pienaar, creator of Feast, teamed up to provide a joint definition of feature stores
and how they can solve the data problem for ML.
*Sponsored post. If you want to be featured here, or as our main sponsor, contact us!

• Data Scientist - Apple Pay Analytics - NYC
  
  You will play a key role improving the Apple Pay product experience. As a member of the analytics team you will be supporting a product function. You will partner with business owners, understand goals, craft KPIs and measure ongoing performance. You will initially engage with the product and engineering teams in ensuring that we have the appropriate instrumentation in place to deliver on these metrics. You will subsequently use advanced statistical, ML and analytical techniques to analyze product performance and identify key insights that inform product improvements and business strategy. The role requires a high degree of independence, ownership and collaboration working cross functionally across all levels of a highly matrixed organization...

        Want to post a job here? Email us for details >> team@datascienceweekly.org

• How to Ace the Case Study Interview as an Analyst
  As a data analyst or data scientist, we not only need to know probabilities and statistics, machine learning algorithms, coding, but most importantly we need to know how to use these techniques to solve any business problems...When I first started to prepare for the case study interview, I didn’t know there are different types of case studies. The fastest way to be an expert in the case study is to know all the frameworks to solve different kinds of case studies...Understanding what the interviewers are looking for can help you better prepare for your answer...

• How to think about correlation? It’s the slope of the regression when x and y have been standardized
  The question is this. Why are correlations meaningful? We know that they are ubiquitous, they get reported all the time in work across many disciplines. But for the life of me I cannot understand what the question is to which a correlation is the answer...

• Variational Autoencoder Demystified With PyTorch Implementation
  This tutorial implements a variational autoencoder for non-black and white images using PyTorch...

• Hands-On Machine Learning with scikit-learn and Scientific Python Toolkits
  Integrate scikit-learn with various tools such as NumPy, pandas, imbalanced-learn, and scikit-surprise and use it to solve real-world machine learning problems...
  For a detailed list of books covering Data Science, Machine Learning, AI and associated programming languages check out our resources page
  .
   

  ---

  
  P.S., Enjoy the newsletter? Please forward it to your friends and colleagues - we'd love to have them onboard :) All the best, Hannah & Sebastian