Subscribe to the Machine Learning Engineer Newsletter

Receive curated articles, tutorials and blog posts from experienced Machine Learning professionals.

Issue #5

This week in Issue #5:
Introducing machine learning jobs, the launch of eXplainableAI v0.0.2, deep neural inspection with DeepBase, AI against Alzheimer's disease, tensorflow differential privacy library, the role of ML in databases and featured libraries on data pipelines!
Support the ML Engineer!
Forward the email, or share the online version on 🐦 Twitter,  💼 Linkedin and  📕 Facebook!
If you would like to suggest articles, ideas, tutorials, libraries or provide feedback just hit reply or send us an email to!
We are extremely excited to announce the launch of XAI v0.0.2 ALPHA, an open source library developed by The Institute for Ethical AI & Machine Learning focused on AI explainability. XAI is currently in very early stages, but we have very exciting plans for this project aligned with our Responsible ML Principle #3. You can install it by running `pip install xai`. You can also learn more about the principles that inspired this project at our Tensorflow London meetup talk.
One of our fellow readers suggested a very interesting paper which was presented at NeurIPS 2018: "Deep Neural Inspection with DeepBase" (Thanks Cecilia Shao from CometML!). In this paper they propose a declarative system for developers to inspect and understand neural networks. They also provide a great blog post where they provide 1) the definition of "deep neural inspection", 2) The inspection of a PacMan Agent with DeepBase, and 3) a link to all contributors. We would like to also point our readers to another very interesting approach to inspect DNNs presented in our 3rd newsletter edition, namely SHAP's DeepLift approach towards explaining predictions.
A research team from the UCSF released a paper which showed a deep learning model that achieved 100% sensitivity (i.e. no false negatives) and 82% specificity (i.e. low false positives) on an average of 6 years prior to final diagnosis. A lot of media articles have also spread the word on this piece. Imaging has been an area in medicine that deep learning has been continuously excelling on, very exciting to see what 2019 has ahead, especially with the current breakthroughs in NLP.
In brief, differential privacy is a technique that allows for a variable level of statistical noise to be added to a dataset. This makes it harder for re-identification attacks (making sure privacy is protected), whilst keeping the statistical properties of the data for analysis to be carried out. This is awesome as it ensures data is not available in its raw form when not necessary, but can still be used for machine learning (Apple provides a great definition). The Tensorflow team has created a new open source library that includes implementations of TensorFlow optimizers for training machine learning models with differential privacy.
Incredibly interesting article by UC Berkeley's Rise Lab on the ML and databases discussion, as well as their recently updated paper, "Learning to Optimize Join Queries With Deep Reinforcement Learning”. In this post they cover how they integrated a form of Deep Q-Learning approach into full-featured query optimizers with minimal modification, achieving signfiicant results.
MLOps = ML Operations
This week's edition is focused on ETL and data pipeline frameworks which fall on our Responsibel ML Principle #4. The four featured libraries this week from the Awesome MLOps list are:
  • Apache Airflow - Data Pipeline framework built in Python, including scheduler, DAG definition and a UI for visualisation
  • Luigi - Luigi is a Python module that helps you build complex pipelines of batch jobs, handling dependency resolution, workflow management, visualisation, etc
  • Genie - Job orchestration engine to interface and trigger the execution of jobs from Hadoop-based systems
  • Oozie - Workflow scheduler for Hadoop jobs
If you know of any libraries that are not in the "Awesome MLOps" list, please do give us a heads up or feel free to add a pull request
Machine Learning Jobs
From this week on, we will be showcasing Machine Learning Engineering jobs (primarily in London for now) to help our community to stay up to date with great opportunities that come up.
It's worth mentioning that these are not sponsored posts. If you have a job you would like us to feature please send us an email, we would especially like to support startups and scale-ups!
Junior Opportunities
Mid-level Opportunities
Leadership Opportunities