Troy H

Engineer & obsessive problem solver, philosopher, inventer & autodidact. Lover of coffee, travel and all things artisan. Currently exploring machine learning.

[Troy H @TroyHolds]

@PerthMLGroup @TTN_Perth this looks promising!

[Pete Warden @petewarden]

"Sparse Architecture Search for CNNs on Resource-Constrained Microcontrollers": https://t.co/lOtWpTgvKP - fascinating work from Arm Research, running models with less than 2KB of memory on embedded systems.

[Troy H @TroyHolds]

@fishnets88 Nice! The possibilities... https://t.co/mfQX5pTt1P

[Matplotlib @matplotlib]

Better dates, smarter legends for scatter plots, arbitrary scales - these are just a few of the features & improvements in our 3.1 release! 🎇Check it out at
https://t.co/QUGTrqBMLg

[Troy H @TroyHolds]

@mbostock @observablehq @mbostock not just code; imagine the benefits of visualising logic in legal contracts, insurance policies, government spending...

[Jeremy Howard @jeremyphoward]

I'm rather blown away by these beautiful SGD visualizations from inspiring @fastdotai participant @ideami - thank you for sharing! :)
https://t.co/E6ctOuhA6i

[Troy H @TroyHolds]

@junpenglao So, are you suggesting that Chemists aren't scientists and that they don't experiment with chemicals?

[Troy H @TroyHolds]

@PerthMLGroup

[Jascha Sohl-Dickstein @jaschasd]

Eliminating All Bad Local Minima from Loss Landscapes Without Even Adding an Extra Unit https://t.co/Wf9Wi0WpHv
It's less than one page. It may be deep. It may be trivial. It will definitely help you understand how some claims in recent theory papers could possibly be true.

[Zaid زيد @zaidalyafeai]

These were my best courses I took in Machine Learning in no particular order

* Coursera Machine Learning by @AndrewYNg
* CS231n course from Stanford by @karpathy and @jcjohnss
* Fastai courses on practical machine learning by @jeremyphoward

What about yours ?

[Troy H @TroyHolds]

@PerthMLGroup

[Wes McKinney @wesmckinn]

The popularity of deep learning frameworks is resulting in more data engineering work in Python, but the library support falls way short. The number of people actively working on data engineering technology for Python (e.g. @ApacheArrow, @ApacheParquet) is shockingly small

[Troy H @TroyHolds]

@PerthMLGroup

[Dustin Tran @dustinvtran]

"Simple, Distributed, and Accelerated Probabilistic Programming". The #NIPS2018 paper for Edward2. Scaling probabilistic programs to 512 TPUv2 cores and 100+ million parameter models. https://t.co/v1z4sJbVAf https://t.co/4h7Lt0zN69

[Troy H @TroyHolds]

@NateSilver538 Asking whether the model is wrong is perhaps the wrong question - they are all wrong. Historical performance is an indicator of how useful the model is. Usefulness is subjective. If performance over many predictions is our concern then 80% correct predictions is probably useful.

[Troy H @TroyHolds]

@PerthMLGroup

[Adam Roberts @ada_rob]

Introducing the MAESTRO dataset (172 hours of aligned MIDI and audio piano performances) and Wave2Midi2Wave: a factorized approach to modeling and generating piano performances.

w/ @fjord41 Andriy Stasyuk @iansimon @huangcza @sedielem

Read 👉 https://t.co/Iz6GKbRulQ
Listen 👇

[Troy H @TroyHolds]

Wondering if there is anyone else out there using #python and #pywin32 to automate @AspenTech software? Particularly interest to hear of sucesses with HYSYS and Flare System Analyzer! #Engineering #Automation #DataScience #digitaltransformation

[Troy H @TroyHolds]

@PerthMLGroup

[Pete Warden @petewarden]

Machine learning in 2KB of RAM: https://t.co/L5XGRL5Z9n

[Troy H @TroyHolds]

Well said! I always try to keep in mind that all models are wrong, but some are useful.

[Sudeep Pillai @spillai]

Really excited to show what the ML team at @ToyotaResearch has been up to lately. Stay tuned, there's more to come in the next few months!

[Yann LeCun @ylecun]

PyTorch 1.0 is here!

https://t.co/Kr0UQq16EF https://t.co/Kr0UQq16EF

[Delip Rao e/σ @deliprao]

GPyTorch: fast Gaussian processes using PyTorch. Available via pip and can be used with regular NNs or independently. Useful tool for characterizing uncertainties with NN predictions.

[Troy H @TroyHolds]

@PerthMLGroup

[hardmaru @hardmaru]

Use your brain as the encoder, and a deep neural network as the decoder. 🧠