hcarlens

6 ms·

1.
AI at IMO 2025: a round-up(xenaproject.wordpress.com)
3 points by hcarlens 1y ago|0 comments
2.
State of ML Competitions (Feb 2025)(mlcontests.com)
1 points by hcarlens 2y ago|0 comments
3.
DeCAF: "the OG foundation model in vision" (ICML 2024)(mlcontests.com)
2 points by hcarlens 2y ago|0 comments
4.
by hcarlens 3y ago
Interesting, thanks for sharing!
5.
by hcarlens 3y ago
That was true in the first Makridakis competition ("M1") in 1982, and possibly until M4 in 2018, but both M5 and M6 were won by what would generally be considered relatively sophisticated methods (e.g. LightGBM). The Wikipedia art
6.
by hcarlens 3y ago
XGBoost, LightGBM, and Catboost are all used quite frequently in competitions. LightGBM is actually marginally more popular than the other two now, but it's pretty close. In the M5 forecasting competition a few years back, many of the
7.
by hcarlens 3y ago
Yeah, this book was incredible and the tech in it has aged extremely well. Have you tried any of Ted Chiang's books? They're also great hard sci-fi. Another one that plays with similar ideas to Permutation City is the Bobiverse s
8.
by hcarlens 3y ago
Also relevant: https://aimoprize.com/ ($10m prize for models that can perform well at IMO)
9.
by hcarlens 3y ago
Agreed, and not only do they not compare their model to Phi-2 directly, the benchmarks they report don't overlap with the ones in the Phi-2 post[1], making it hard for a third party to compare without running benchmarks themselves. (In
10.
My AI Timelines Have Sped Up (Again)(alexirpan.com)
50 points by hcarlens 3y ago|95 comments
11.
by hcarlens 3y ago
There have been quite a few interesting Kaggle competitions in recent years, as well as other interesting ML/data science competitions on other platforms. Platforms like Kaggle, DrivenData, Zindi, AIcrowd, CodaLab and others are runnin
12.
by hcarlens 3y ago
An interesting approach I came across at NeurIPS a few weeks ago is called "ML with Requirements"[1]: https://arxiv.org/abs/2304.03674 My basic understanding is that it combines "standard" supervise
13.
NeurIPS 2023: Day 1(mlcontests.com)
1 points by hcarlens 3y ago|0 comments
14.
by hcarlens 3y ago
I had a chance to chat to them again today, and wrote some more details here: https://mlcontests.com/neurips-2023/tutorials/#exhibit-hall Also as the other comment mentioned, https://positron.ai seems
15.
NeurIPS 2023: Expo Day(mlcontests.com)
70 points by hcarlens 3y ago|12 comments
16.
by hcarlens 3y ago
I built a slightly less detailed version of this, which also also lists free credits: https://cloud-gpus.com/ Open to any feedback/suggestions! Will be adding 4090/H100 shortly.
17.
by hcarlens 3y ago
Having your own GPU is nice, but it's power hungry, heats the room, and only really makes sense if you use it a lot. For a list of cloud GPU providers with rough price comparisons, I created this page: https://cloud-gpus.com
18.
by hcarlens 4y ago
If anyone is interested in a GPU-specific comparison across these three and other providers, I created this page for that: https://cloud-gpus.com Happy to get any feedback on this!
19.
by hcarlens 4y ago
Good try, and there are indeed different types of infinities, but your examples both represent the same "countable" infinity of the natural numbers! You can create a bijection between these two sets. I can't see anything help
20.
Cloud GPUs: AWS vs. Google vs. Others(cloud-gpus.com)
1 points by hcarlens 5y ago|1 comments
21.
How to Win at Kaggle (2022)(blog.mlcontests.com)
2 points by hcarlens 5y ago|0 comments
22.
An analysis of competitive ML in 2021(blog.mlcontests.com)
2 points by hcarlens 5y ago|0 comments
23.
Top Cloud GPU Providers for ML (2022)(medium.com)
1 points by hcarlens 5y ago|0 comments
24.
by hcarlens 5y ago
Hi! I started mlcontests.com because I wanted to be able to have a simple overview of live Kaggle-style competitions, and I've been running it as a side project for almost two years now. Would love to hear your thoughts/suggestion
25.
Show HN: Discover Ongoing ML Competitions(mlcontests.com)
2 points by hcarlens 5y ago|1 comments
26.
by hcarlens 5y ago
Yeah it's a shame. I don't own the IP, my previous employer does, and I think when they got it published in the journal of trading they had to agree to some level of exclusivity which meant it can't just be uploaded elsewhere
27.
by hcarlens 5y ago
The ones that are currently live tend to have features that try to negate this (randomised uncrossing times, price collars, order priority based on arrival time or size)
28.
by hcarlens 5y ago
Haha yeah! Though times for these auctions are double-digit milliseconds, a lifetime for your fpga strategies! And it's still fairly niche, these are complementing CLOBs/dark pools rather than replacing them. Where did you move t
29.
by hcarlens 5y ago
When they were first introduced, each of the fba books had slightly different mechanics (matching priority, timing, price determination), so each book needs a slightly different approach. I guess you could see it as gamesmanship, but in equ
30.
by hcarlens 5y ago
Actually, it got implemented by most major exchanges in Europe in 2017/2018 and cboe seems to be bringing this to the US now: https://www.cboe.com/us/equities/trading/offerings/periodic_...

More ›