Sep 27, 2019 · newsletter

Theories of (machine) learning should shape horizon scanning - not just application

And no, not this kind of horizon… (image credit)

In a recent newsletter, Alice mused about how evolving views and theories of learning are shaping machine learning research and practice. If you’re an enterprise data scientist you’re very much focused on the practice of machine learning. Limited awareness of what’s shaping the machine learning breakthroughs that you’re trying to apply to real-life problems makes it easy to get locked into certain approaches and to inherit blind spots. The value of this type of epistemological insight to data science practitioners is obvious, but why should business leaders care about it, too?

One good reason is that it makes for more discerning consumption of news about machine learning developments, and this, in turn, sharpens business leaders’ horizon scanning capability. It’s useful to know that there isn’t wholesale consensus on the most fruitful path to advancing machine learning. The whole point of horizon scanning is to work out the best way to help your organization prepare and benefit from the broader trends, whether they are driven by technology, government policy or shifting consumer tastes. This, in turn, directly shapes the level and pace of your investment in your applied machine learning environment - specifically: talent, infrastructure, and governance. It should also guide your decisions about when and how to go about integrating machine learning into your business operations.

Digging into what are tech companies are doing as well as thinking

Several of the thought leaders referenced in Alice’s article are based in large tech companies. These organizations are increasingly sources of cutting edge machine learning research. And while the scale of industry-driven research has raised some valid concerns, there are some very real benefits to businesses:

  • First, industry-driven research affords business leaders the opportunity to make informed bets about the broad viability of a particular approach to machine learning (beyond the soundness of the technology) before deciding to invest heavily themselves. For example, given Google’s and Facebook’s heavy investments in deep learning, it’s reasonable to bet on the emergence of a number of open source data science tools to facilitate this approach. This matters because access to these types of tools is critical for impactful and cost-effective data science practice within enterprises - and building them is notoriously hard to do. The availability of a broad range of open source tools (and libraries) maintained and supported by an enthusiastic and committed community is an important point of consideration when thinking about how to reliably and affordably build out your machine learning infrastructure planning. This is true even if you are swayed by the views of people such as Gary Marcus, who has expressed skepticism about deep learning as the sole path towards artificial general intelligence, and instead advocates a hybrid approach that incorporates not just supervised forms of deep learning, but also techniques such as symbol-manipulation and (a re-conceptualized form of) unsupervised learning.

  • Second, the publicly accessible accounts of how these tech companies are implementing the fruit of their research provides a window into the practical challenges of trying to integrate machine learning into more and more of their day-to-day business operations, and doing so at scale. From a horizon scanning perspective, this is a rich source for useful insights into what could go wrong and what appears to work well. Further, because some of these organizations publish both accounts of their research and also of their (business) implementation efforts, you can read others’ assessments of both of these things. These external reviews/commentary lack the rigor of academic peer reviews, but they do provide useful alternative perspectives for business leaders looking to assess the merits or demerits of a particular approach to applied machine learning. Take for example, Gary Marcus’s view of DeepMind’s framing of its systems as capable of “autonomously learn[ing] how to model other agents in its world” as misleading. In a blog post on DeepMind’s Machine Theory of Mind paper, he remarks, “in fact, in close parallel with the Go work, DeepMind has quietly built in a wealth of prior assumptions.”

You may disagree with his assessment of reinforcement learning - or may not be certain what to think about it - but if reinforcement learning is something your organization is considering, then his comments highlight implementation considerations that are worth discussing with your in-house technical team.

And, as noted above, that’s the whole point of horizon scanning - trying to predict the challenges and opportunities that lie ahead, and trying to work out how best to navigate them before they become a reality. Asking good questions that lead to good conversations - with both your ML practitioners and operation teams - is a good way of doing this.

If you want to dedicate more time to understanding the current machine learning landscape, but aren’t sure where to start, I recommend Adrian Colyer’s summaries of the Microsoft, Facebook and Google papers on their implementation efforts as a good place to start. I think he’s done a good job of capturing the most interesting points in those papers, and he presents them in an accessible way.

Read more

Oct 30, 2019 · newsletter
Sep 27, 2019 · newsletter

Latest posts

Sep 22, 2021 · post

Automatic Summarization from TextRank to Transformers

by Melanie Beck · Automatic summarization is a task in which a machine distills a large amount of data into a subset (the summary) that retains the most relevant and important information from the whole. While traditionally applied to text, automatic summarization can include other formats such as images or audio. In this article we’ll cover the main approaches to automatic text summarization, talk about what makes for a good summary, and introduce Summarize. – a summarization prototype we built that showcases several automatic summarization techniques. more
Sep 21, 2021 · post

Extractive Summarization with Sentence-BERT

by Victor Dibia · In extractive summarization, the task is to identify a subset of text (e.g., sentences) from a document that can then be assembled into a summary. Overall, we can treat extractive summarization as a recommendation problem. That is, given a query, recommend a set of sentences that are relevant. The query here is the document, relevance is a measure of whether a given sentence belongs in the document summary. How we go about obtaining this measure of relevance varies (a common dilemma for any recommendation system). more
Sep 20, 2021 · post

How (and when) to enable early stopping for Gensim's Word2Vec

by Melanie Beck · The Gensim library is a staple of the NLP stack. While it primarily focuses on topic modeling and similarity for documents, it also supports several word embedding algorithms, including what is likely the best-known implementation of Word2Vec. Word embedding models like Word2Vec use unlabeled data to learn vector representations for each token in a corpus. These embeddings can then be used as features in myriad downstream tasks such as classification, clustering, or recommendation systems. more
Jul 7, 2021 · post

Exploring Multi-Objective Hyperparameter Optimization

By Chris and Melanie. The machine learning life cycle is more than data + model = API. We know there is a wealth of subtlety and finesse involved in data cleaning and feature engineering. In the same vein, there is more to model-building than feeding data in and reading off a prediction. ML model building requires thoughtfulness both in terms of which metric to optimize for a given problem, and how best to optimize your model for that metric! more
Jun 9, 2021 ·

Deep Metric Learning for Signature Verification

By Victor and Andrew. TLDR; This post provides an overview of metric learning loss functions (constrastive, triplet, quadruplet, and group loss), and results from applying contrastive and triplet loss to the task of signature verification. A complete list of the posts in this series is outlined below: Pretrained Models as Baselines for Signature Verification -- Part 1: Deep Learning for Automatic Offline Signature Verification: An Introduction Part 2: Pretrained Models as Baselines for Signature Verification Part 3: Deep Metric Learning for Signature Verification In our previous blog post, we discussed how pretrained models can serve as strong baselines for the task of signature verification. more
May 27, 2021 · post

Pretrained Models as a Strong Baseline for Automatic Signature Verification

By Victor and Andrew. Figure 1. Baseline approach for automatic signature verification using pretrained models TLDR; This post describes how pretrained image classification models can be used as strong baselines for the task of signature verification. The full list of posts in the series is outlined below: Pretrained Models as Baselines for Signature Verification -- Part 1: Deep Learning for Automatic Offline Signature Verification: An Introduction Part 2: Pretrained Models as Baselines for Signature Verification Part 3: Deep Metric Learning for Signature Verification As discussed in our introductory blog post, offline signature verification is a biometric verification task that aims to discriminate between genuine and forged samples of handwritten signatures. 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