engblogs

Using OCaml to implement deep reinforcement learning experiments.

An exploration of the interaction between L2 regularization and batch normalization in machine learning.

Explores implementing a custom Redux garbage collector that tracks on-screen selectors and sweeps unused entities to fit memory constraints in a React-Redux Xbox app.

Case study of a memory-conscious Redux garbage collector for a React/Redux app (SoundCloud Xbox), leveraging onscreen selectors, a normalized state store, and a sweep phase to drop unused entities without impacting performance.

Garbage collection in Redux applications on the SoundCloud Xbox application running on Microsoft's UWP framework.

A tutorial on building web applications with Incr_dom, an in-house framework at Jane Street, modeled after React's virtual DOM and built on top of Incremental for efficient rendering.

Highlights of the OpenAI Fellows Summer 2018 final projects

Exploring the scalability of AI training methods in depth

Exploring Capstone, a tablet that facilitates a two-step creative process through the use of cards and inking on a freeform canvas.

Exploring the quantification of generalization in reinforcement learning.

Using integration testing with a ModuleFactory-based VIPER module to detect and prevent memory leaks in a Swift iOS app.

Using integration testing to detect and prevent memory leaks in VIPER-based Swift modules by validating module lifecycles via a ModuleFactory and inspecting weak versus strong references.