Apr 3, 2019 · post

An Invitation to Active Learning

Many interesting learning problems exist in places where labeled data is limited. As such, much thought has been spent on how best to learn from limited labeled data. One obvious answer is simply to collect more data. That is valid, but for some applications, data is difficult or expensive to collect. If we will collect more data, we ought at least be smart about the data we collect. This motivates active learning, which provides strategies for learning in this scenario.

The ideal setting for active learning is that in which we have a small amount of labeled data with which to build a model and access to a large pool of unlabeled data. We must also have the means to label some of that data, but it’s OK for the labeling process to be costly (for instance, a human hand-labeling an image). The active learning process forms a loop:

  1. build a model based on the labeled data available
  2. use the model to predict labels for the unlabeled points
  3. use an active learning strategy to decide which point to label next
  4. label that point
  5. GOTO 1.

The active learning loop in action The active learning loop in action - try out the demo!

The essence of active learning is in the strategy we choose in the loop above. Three broad families of strategy are:

  • Random sampling. In the default case, we sample unlabeled data from the pool randomly. This is a passive approach where we don’t use the output of the current model to inform the next data point to be labeled. As such, it isn’t really active learning.

  • Uncertainty sampling. In uncertainty sampling, we choose the data point about which the algorithm is least certain to label next. This could be the point closest to the decision boundary (the least confident prediction), or it could be the point with highest entropy, or other measure of uncertainty. Choosing points as such helps our learning algorithm refine the decision boundary.

  • Density sampling. Uncertainty sampling works much better than random sampling, but by definition it causes the data points we choose to label to cluster around the decision boundary. This data may be very informative, but not necessarily representative. In density sampling, we try to sample from regions where there are many data points. The trade off between informativeness and representativeness is fundamental to active learning, and there are many approaches that address it.

To illustrate the difference between passive and active learning, we created an Observable notebook with a toy problem, which you can explore here. In the notebook, the goal is to find a good separation of red and blue points on a two dimensional chart, and we train a logistic regression model live in the browser to do so. One can see how the decision boundary separating the points evolves as more data is labeled with a random sampling strategy, and also with an uncertainty sampling strategy. In the case of two classes with a linear decision boundary, all the uncertainty sampling strategies (least confidence and highest entropy) give the same result. This is an extremely simplified example, but we think it shows some of the intuition behind active learning.

We explore active learning in much more detail in our report Learning with Limited Labeled Data, and you can get a high level overview in our previous post.

Read more

Apr 29, 2019 · newsletter
Apr 2, 2019 · featured post

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