Tim Anderton

machine learning for fun and profit (mostly fun).

[Demetri (is over at the other place too) @PhDemetri]

Bro, if you were to find a job using y=mx+b, I am fairly certain you could double your income. No joke

[Vala Afshar @ValaAfshar]

This is how we should teach physics and engineering principles in high school

[Pedro Domingos @pmddomingos]

All of linear algebra in one diagram:

[Demetri (is over at the other place too) @PhDemetri]

I’m gonna write a book on stats one day, and there will be an entire chapter on not being such a rule based hard ass and knowing when to commit a statistical sin in the name of getting an approximate answer to a precise question

[Gabriel Peyré @gabrielpeyre]

Gradient descent/ascent can be inefficient to find saddle point (Nash equilibrium) for min-max games, because of spiralling behaviour. https://t.co/KKKGQ48yl0

[François Chollet @fchollet]

Tip #1: when building models with the Functional API (https://t.co/suUPf0h2Zt), you can use an intermediate output as entry point to define a new Model.

This is especially useful to create pairs of models where one includes preprocessing (e.g. for inference), and one does not.

[Demetri (is over at the other place too) @PhDemetri]

I believe eigen fly

I believe eigen touch the sky

[François Chollet @fchollet]

The dumbest, most impotent idea of the 2010s and 2020s is the notion that you can fight disinformation with actual information. It's like fighting a forest fire by planting trees.

[François Chollet @fchollet]

Wait, what

[Tim Anderton @anderton_tim]

Really looking forward to this Boston cream Pi :)

[Andrej Karpathy @karpathy]

Everyone is so obsessed with accelerating neural nets, so as a fun side project I've been building this breadboard 8bit neural net decelerator. It will crawl at best :D. (following along the excellent++ Ben Eater 8-bit computer https://t.co/iDw8gqNnGT)

[Peter Wang 🦋 @pwang]

Just a reminder that there is no such thing as data, only frozen models.

I’m not particularly encouraged by the models being frozen into these metrics.

[Josh Greenberg @epistemographer]

Wanted: a group videoconference interface that includes a running "chess clock" indicating the proportion of the event time that each participant has talked.

[Jesse Engel @jesseengel]

Wouldn't it just be a lot more self explanatory is we called Transformers something like Key-Query Networks (KQNets)?

[Francis Bach @BachFrancis]

Nothing beats some polynomial magic to start the week, with a new blog post on Chebyshev polynomials. https://t.co/iI1xxTOdl2

[Lex Fridman @lexfridman]

New proposal for Turing Test in robotics: The Door Handle & Water Tub Test. The task is to achieve cat-level perception, reasoning, learning, manipulation, and dynamic planning under uncertainty.

[hardmaru @hardmaru]

Using gradient descent for everything https://t.co/okUR4RKTiQ

[Brian P. Hogan @bphogan]

I’ve hired a few people in my time.

I’ve never been disappointed by hiring someone who is less technical but is a great human who cares for others.

I can teach tech pretty easily.

But boy have I seen people who are strong technical people with toxic attitudes destroy a team.

[Catherine Olsson @catherineols]

Wow, what a plot twist in the abstract! (@sandsubramanian et al)

[Hiten Shah @hnshah]

Happy Monday!

Inspiration for the week. Sometimes it takes a few attempts to get through a seemingly impossible barrier.

https://t.co/G1ObfDNd0k