Saki Shinoda by

Data-Efficient Deep Learning

2018-03-20 22:45:00 +0000

I had the privilege of presenting a brief overview of tools for data-efficient deep learning at WiDS Zurich. The presentation slides are available for download here.

Virtual Adversarial Ladder Networks for Semi-Supervised Learning

2017-11-28 18:49:00 +0000

Authors: Saki Shinoda, Daniel E. Worrall, Gabriel J. Brostow

Semi-supervised deep learning classification results

2017-07-05 15:27:00 +0000

Best self-reported performances on semi-supervised image classification benchmarks. Inspired by Rodrigo Benenson’s mini-site collating published performance on key computer vision datasets.

Approaches to running Tensorflow on AWS

2017-02-23 13:09:00 +0000

Work in progress

GPU-enabled Tensorflow in Docker on AWS

2017-02-13 02:22:00 +0000

Update (again): Not described here are the security group settings necessary for accessing the Jupyter notebook running on the AWS instance (or SSH-ing in). Again, something I might add in future.

Tensorflow summary event files as numpy arrays

2017-02-13 01:58:00 +0000

Pass this function a directory path, and it returns the path to the most recently modified file in that directory. This can be used to get the most recent event file in your logdir.

Graph visualization

2017-02-13 01:14:00 +0000

There was going to be a series of posts about how to make pretty(-ish) graphs using NetworkX in Python and VisNetwork in R. I started working on a few back in July. However, I don’t really have time to condense all the various notes I made while working at NIBR. So instead, a few bits of miscellany follow and may be added to in the future.

Notes on building this site

2017-02-12 21:47:00 +0000

I thought jekyll and Github Pages would be an easy way to get a website up and running with minimal CSS and HTML hackery. Like many a millennial I learned CSS and HTML at a very early age, but I don’t much find it particularly fulfilling to wrangle with these days. What has been a little more interesting is learning about Ruby (I guess).

Rescue lost code in Jupyter/IPython notebooks

2016-07-24 10:00:00 +0000

I found this in a moment of great need here. It explains it far better than I do so go check it out. In short though, to rescue a function that is still in your environment, but that you defined in a cell that you’ve since deleted, use the following nifty little function.

def rescue_code(function):
    import inspect
    get_ipython().set_next_input("".join(inspect.getsourcelines(function)[0]))