machine learning for fun and profit (mostly fun).
Joined October 2014
- Tweets 90
- Following 279
- Followers 30
- Likes 53
5 Photos and videos
Tim Anderton retweeted
Bro, if you were to find a job using y=mx b, I am fairly certain you could double your income. No joke
5
4
94
Tim Anderton retweeted
20 Aug 2022
This is how we should teach physics and engineering principles in high school
147
3,734
20,063
Tim Anderton retweeted
10 Aug 2022
All of linear algebra in one diagram:
61
1,190
6,230
Tim Anderton retweeted
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
2
4
56
Tim Anderton retweeted
18 Mar 2022
Gradient descent/ascent can be inefficient to find saddle point (Nash equilibrium) for min-max games, because of spiralling behaviour. en.wikipedia.org/wiki/Nash_e…
8
176
1,000
Tim Anderton retweeted
12 Mar 2022
Tip #1: when building models with the Functional API (keras.io/guides/functional_a…), 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.
5
12
133
Tim Anderton retweeted
I believe eigen fly
I believe eigen touch the sky
56
998
7,472
Tim Anderton retweeted
21 Aug 2021
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.
12
49
284
Tim Anderton retweeted
7 Feb 2021
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 eater.net/8bit/)
45
119
1,716
Tim Anderton retweeted
30 Nov 2020
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.
24 Nov 2020
Esoteric metrics based on analyzing extensive data about employee activities has been mostly the domain of fringe software vendors. Now it's built into MS 365.
A new feature to calculate 'productivity scores' turns Microsoft 365 into an full-fledged workplace surveillance tool:
1
2
6
Tim Anderton retweeted
27 Apr 2020
Wanted: a group videoconference interface that includes a running "chess clock" indicating the proportion of the event time that each participant has talked.
7
21
131
Tim Anderton retweeted
26 Nov 2019
Wouldn't it just be a lot more self explanatory is we called Transformers something like Key-Query Networks (KQNets)?
9
4
118
Tim Anderton retweeted
4 Nov 2019
Nothing beats some polynomial magic to start the week, with a new blog post on Chebyshev polynomials. francisbach.com/chebyshev-po…
5
87
388
Tim Anderton retweeted
25 Sep 2019
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.
37
340
2,139
Using gradient descent for everything x.com/danhett/status/1176116…
41
422
2,351
Tim Anderton retweeted
15 Sep 2019
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.
507
11,762
56,668
Tim Anderton retweeted
10 Sep 2019
Wow, what a plot twist in the abstract! (@sandsubramanian et al)
44
1,077
3,323
Tim Anderton retweeted
9 Sep 2019
Happy Monday!
Inspiration for the week. Sometimes it takes a few attempts to get through a seemingly impossible barrier.
x.com/nosabesnadajon/status/…
253
4,595
24,602