eggie5

19 ms·

181.
by eggie5 9y ago
the post wasn't about arkit, it was about coreml.
182.
by eggie5 9y ago
I'm just super jealous every time I see they pay 10k over market on average.
183.
by eggie5 9y ago
there goes booking.com ruining the std deviation in Amsterdam again! :)
184.
by eggie5 9y ago
SVD is the decomposition of a matrix into its components. PCA is the analysis of a set of eigenvectors. Eigenvectors can come from SVD components or a covariance matrix. source: http://www.eggie5.com/107-svd-and-pca
185.
by eggie5 9y ago
most of the top submissions to the last ImageNet comp. are tweaks on YOLO and Faster RCNN. But why is this trending now?
186.
by eggie5 9y ago
Yes they are very smelly
187.
by eggie5 9y ago
I've always liked this visualization of the Bias-Variance tradeoff: http://www.eggie5.com/110-bias-variance-tradeoff
188.
Nightmare Eigenface(eggie5.com)
2 points by eggie5 9y ago|0 comments
189.
by eggie5 9y ago
I remember I made this same thing when I was in undergrad b/c my uni didn't have waitlists. So people could waitlist on my app. Their priority in the waitlist is determined by a market. I remember getting SMS's in the middle
190.
by eggie5 9y ago
No, it was an ensemble of Collaborative filtering using Matrix Factorization (using SVD) and a RBM.
191.
by eggie5 9y ago
just when I have everything migrated to slack...
192.
Bias-Variance Tradeoff(eggie5.com)
3 points by eggie5 9y ago|0 comments
193.
by eggie5 9y ago
types in python 3.5?! I had no idea -- that's exciting.
194.
by eggie5 9y ago
You can find pre-trained models on the website of popular websites like caffe, theano and tensorflow in a section that is typically called a "model zoo".
195.
SVD and PCA: connections?(eggie5.com)
1 points by eggie5 9y ago|0 comments
196.
by eggie5 9y ago
yes, I would give up Highlights, Threads, Calls and most slack features for a native client.
197.
by eggie5 9y ago
They use sampling to "preserve a fixed ratio of positive to negative examples." I wonder what the ratio is... isn't the technique w/ class imbalances to up-sample/subsample to get a 50/50 ratio so your classifi
198.
by eggie5 9y ago
What modern server is this describing? "a number of threads ... which poll message queues..." and what is this describing? "...a completely independent thread for every event." "Languages based around message passi
199.
by eggie5 9y ago
it is a very interesting idea, considering how GD updates are inherently a serial operation. I've always wondered how they pulled it off in spark... I've also read about hogwild which allows parallel SGD on sparse datasets...
200.
Ask HN: Istio vs. Envoy
1 points by eggie5 9y ago|0 comments
201.
by eggie5 9y ago
It would be interesting to quantify this. I did to go through the rigmarole of building tensorflow for GPU training in GCE (building TF: installing nvidia drivers, cudnn, etc) and docker would have definitely been a boon! Or it would be nic
202.
by eggie5 9y ago
Although fine for your relative benchmarking, do you notice that docker introduces noticeable overhead to training time vs non-docker?
203.
by eggie5 9y ago
Looking forward to your model deployment blog post -- it's still a new pattern for most
204.
by eggie5 9y ago
That example is essentially building (and ostensibly training) an albeit a trivial model in Go. Typically you have more complicated architecture so you're deployment has two parts: * the weights * and the architecture
205.
by eggie5 9y ago
I wish the diagrams were bigger, they are hard to read and a bit blurry. One of the interesting points, that is often overlooked in ML is model deployment. They mention tensorflow, which has a model export feature that you can use as long a
206.
by eggie5 9y ago
probably a lot of selection bias on these responses.
207.
by eggie5 9y ago
I don't think the two are comparable. Django is a full featured and opinionated MVC web framework, while Flask is a much more simple library to handle web apps. I think flask has it's place for narrow-scoped projects like some typ
208.
by eggie5 9y ago
Prototype with Keras and then native TensorFlow if I want to fine tune.
209.
by eggie5 9y ago
I think he's talking about libsvm, the c++ package
210.
by eggie5 9y ago
oh, you're talking about updating parameters. Maybe sending that new data home and train a new model and then update parameters across all devices in fleet.

More ›