Google's Hybrid Approach to Research - Spector et al. 2012 Something a little different to close out the week, a paper describing how Google conduct research. It's a fascinating look at how they balance fundamental and applied research, how they integrate research into product teams, and how they measure the contribution of the research. I … Continue reading Google’s Hybrid Approach to Research
Inferring Causal Impact Using Bayesian Structural Time-Series Models
Inferring Causal Impact Using Bayesian Structural Time-Series Models - Brodersen et al. (Google) 2015 Today's paper comes from 'The Annals of Applied Statistics' - not one of my usual sources (!), and a good indication that I'm likely to be well out of my depth again for parts of it. Nevertheless, it addresses a really … Continue reading Inferring Causal Impact Using Bayesian Structural Time-Series Models
Graying the Black Box: Understanding DQNs
Graying the Black Box: Understanding DQNs - Zahavy et al. 2016 It's hard to escape the excitement around deep learning these days. Over the last couple of days we looked at some of the lessons learned by Google's machine learning systems teams, including the need to develop ways of getting insights into the predictions made … Continue reading Graying the Black Box: Understanding DQNs
Ad Click Prediction: A View from the Trenches
Ad Click Prediction: a View from the Trenches - McMahan et al. 2013 Yesterday we looked at a tour through the many ways technical debt can creep into machine learning systems. In that paper, the authors mention an automated feature management tool that since its adoption, "has regularly allowed a team at Google to safely … Continue reading Ad Click Prediction: A View from the Trenches
Machine Learning: The High-Interest Credit Card of Technical Debt
Machine Learning: The High-Interest Credit Card of Technical Debt - Sculley et al. 2014 Today's paper offers some pragmatic advice for the developers and maintainers of machine learning systems in production. It's easy to rush out version 1.0 the authors warn us, but making subsequent improvements can be unexpectedly difficult. You very much get the … Continue reading Machine Learning: The High-Interest Credit Card of Technical Debt
Distributed Consistency and Session Anomalies
Since we've spent the last couple of days sketching anomaly diagrams and looking at isolation levels, I wanted to finish the week off with a quick recap of session anomalies and consistency levels for distributed stores. In terms of papers, I've drawn primary material for this from: Highly Available Transactions: Virtues and Limitations, and Linearizability … Continue reading Distributed Consistency and Session Anomalies
Generalized Isolation Level Definitions
Generalized Isolation Level Definitions - Adya et al. 2000 Following on from yesterday's critique of ANSI SQL isolation levels, today's paper also gives a precise definition of isolation levels - but does so in a way that is inclusive of optimistic and general multi-version concurrency control strategies instead of being dependent on locking. Where Berenson … Continue reading Generalized Isolation Level Definitions
A Critique of ANSI SQL Isolation Levels
A Critique of ANSI SQL Isolation Levels - Berenson et al. 1995 udpate: 2 minor corrections in the section on A5A per the comment from 'banks' - thanks! The ANSI SQL isolation levels were originally defined in prose, in terms of three specific anomalies that they were designed to prevent. Unsurprisingly, it turns out that … Continue reading A Critique of ANSI SQL Isolation Levels
Not-quite-so-broken TLS: lessons in re-engineering a security protocol specification and implementation
Not-quite-so-broken TLS: lessons in re-engineering a security protocol specification and implementation - Kaloper-Meršinjak et al. 2015 Update: fixed broken paper link above. On the surface this is a paper about a TLS implementation, but the really interesting story to me is the attempt to 'do it right,' and the techniques and considerations involved in that … Continue reading Not-quite-so-broken TLS: lessons in re-engineering a security protocol specification and implementation
IncludeOS: A minimal, resource efficient unikernel for cloud systems
IncludeOS: A minimal, resource efficient unikernel for cloud systems - Bratterud et al. 2015 There has been lots of excitement around unikernels over the last year, and especially with the recent acquisition of the Unikernel Systems team by Docker (MirageOS, Mergeable Persistent Data Structures, Jitsu: Just-in time summoning of Unikernels). Whereas MirageOS is built around … Continue reading IncludeOS: A minimal, resource efficient unikernel for cloud systems
the morning paper 