bradneuberg

25 ms·

91.
by bradneuberg 10y ago
See my reply to dnt404 for two examples. We have nearly an exabyte of data at Dropbox ( https://blogs.dropbox.com/tech/2016/03/magic-pocket-infrastr... ); there's lots of interesting ML things we can do
92.
by bradneuberg 10y ago
Two examples of where we use machine learning at Dropbox is corner and edge detection in our mobile docscanner ( https://blogs.dropbox.com/tech/2016/08/fast-and-accurate-doc... ) and computer vision for OCRing.
93.
by bradneuberg 10y ago
No public posts yet unfortunately. For spinning up jobs we have an internal tool that can spin up machines and provision them, but in my own experiments I just usually end up spinning up machines and rsyncing experiments over.
94.
by bradneuberg 10y ago
At Dropbox we use different machines for training and inference, as their computational needs are so different. We train on Amazon GPU machines, while we deploy on our own Dropbox compute infrastructure on CPU machines.
95.
by bradneuberg 10y ago
(ML engineer from Dropbox here) For deep learning oriented projects, we train on EC2 GPU instances, generally p2.xlarge instances these days to get the Nvidia K80 GPUs. We can spin up many of these in parallel if we are doing model architec
96.
by bradneuberg 10y ago
If this becomes common place across borders phones should come with a "fake" unlock access code - if you enter it it drops you into a plain vanilla setup with some fake contacts perhaps. Might make sense to create fake email and s
97.
by bradneuberg 10y ago
Just donated 25 bucks. Octave is a fabulous tool.
98.
by bradneuberg 10y ago
Jürgen Schmidhuber is an early neural net pioneer, especially with regards to making recurrent neural nets a reality. As others have mentioned he was instrumental with LSTMs, but also advised Alex Graves on his work with Connectionist Tempo
99.
by bradneuberg 10y ago
Just gave. John has given so much to all of us, good to give a bit back to him.
100.
by bradneuberg 10y ago
I'd love to learn more about the techniques actually being used in thse systems. Any good pointers to some scientific papers or review articles on the subject? I have a background in machine learning so am comfortable with technical pa
101.
by bradneuberg 10y ago
Is this both automated defense and offense via machine learning, or just automated defensive systems? If it includes automated offensive systems, what's to keep these kinds of systems from jumping outside of their sandboxes and comprom
102.
by bradneuberg 11y ago
"There's been some other papers recently showing the persistence of memories even when synapses and spines are lost (for example, during hibernation)" Interesting; do you have any links to those papers with more information?
103.
by bradneuberg 11y ago
Unless we ourselves are composed of self replicating probes. It's still a bit of a mystery how quickly life appeared on Earth itself. Perhaps Earth was seeded via panspermia, and the cells themselves are remnants of self replicating pr
104.
by bradneuberg 11y ago
NASA had a plan in the early 70s to have humans on the surface of Mars by 1990. When Nixon canceled the Apollo and Saturn five missions, all that early planning went into the trashbin.
105.
Surprising Results of What It Would Take to Transport a Million People to Mars(blog.codinginparadise.org)
1 points by bradneuberg 11y ago|0 comments
106.
by bradneuberg 11y ago
FYI a technical article with a full write up on Cloudless is here: http://codinginparadise.org/ebooks/html/blog/introducing_clo...
107.
by bradneuberg 11y ago
[Cloudless developer here] Hi Boris! It's Brad from Dropbox :) Responding to the parent comment, in terms of running time, the primary bottleneck is the RCNN localization portion; that takes about a minute and a half on my laptop to pr
108.
by bradneuberg 11y ago
The primary challenge is collecting enough training data to teach the neural network.
109.
by bradneuberg 11y ago
[Cloudless developer here] It can certainly be done with hand rolled heuristics pipelines. However, one of the trends in machine learning is what is known as end to end learning - having your machine learning model automatically discover wh
110.
by bradneuberg 11y ago
[Cloudless developer here] I believe the Landsat cover percent field is generated by humans and historical information, but I'm not completely sure.
111.
by bradneuberg 11y ago
[Cloudless developer here] Our backgrounds are more in machine learning and computer science than metereorology, so that aspect was new to us. As you mention we were mostly approaching the problem from having more data and tools like neural
112.
by bradneuberg 11y ago
[Cloudless developer here] Thanks for all the great suggestions! In terms of your existing 98% accuracy solution, can you point me to more details if possible? I grabbed AlexNet as its a bit easier to work with and was readily available as
113.
Cloudless: Open Source Deep Learning Pipeline for Orbital Satellite Data(codinginparadise.org)
5 points by bradneuberg 11y ago|0 comments
114.
by bradneuberg 11y ago
In the paper they use the abbreviation mAP without explaining what it is or providing a reference, such as "Fast YOLO, processes an astounding 155 frames per second while still achieving double the mAP of other real-time detectors&quot
115.
by bradneuberg 11y ago
It's not just ReLUs that have helped, its also better random initialization before starting training, such as using Xavier initialization ( http://andyljones.tumblr.com/post/110998971763/an-explanatio... ) Also
116.
by bradneuberg 11y ago
I completed this course over the last year. It's fantastic and from one of the founders of the field. I'd stick to the first half to get a good sense of back propagation and working with standard neural nets. I'd hold off on
117.
Ten Deep Learning Trends I Saw at NIPS 2015(codinginparadise.org)
3 points by bradneuberg 11y ago|0 comments
118.
NIPS Day 1: TensorFlow and more(codinginparadise.org)
3 points by bradneuberg 11y ago|0 comments
119.
by bradneuberg 11y ago
I bought a Commodore 64 + monitor a few years ago on eBay and played the original Ultima IV on it to completion.
120.
by bradneuberg 11y ago
I grew up on a CoCo3; loved that thing as well :) OS-9 Level II was where I first learned a Unix-like OS (on a tiny 6809 machine!) OS-9 Level II was truly a remarkable achievement for such a tiny machine: multi-tasking, windowing, and more.

More ›