prats226

10 ms·

91.
by prats226 9y ago
I am Prathamesh, co-founder of https://nanonets.com We were working on a project to detect objects using deep learning with raspberry pi and we have benchmarked various deep learning architectures on pi. With ~100-200 images, yo
92.
by prats226 9y ago
For those who don't like to wait for access, https://nanonets.com Just upload your training data and we will provide machine learning API automatically.
93.
NanoNets (YC W17) Is Hiring Founding Data Scientist in San Francisco(angel.co)
1 points by prats226 9y ago
94.
Ask HN: What are some of the heuristics you use for hyperparameter search?
2 points by prats226 9y ago|0 comments
95.
The top 8 companies from Y Combinator W17 Demo Day 2(techcrunch.com)
1 points by prats226 10y ago|0 comments
96.
by prats226 10y ago
What are other providers that you use? Since you are using it as input to your training set, it should have fairly high accuracy (even more accurate than manual errors in tagging)
97.
by prats226 10y ago
What is accuracy of state of the art algorithm for audio captioning? YouTube has about 95% accuracy according to some folks. And what is the accuracy needed by regulation? If cost of running algorithms is the only issue, there should be a s
98.
by prats226 10y ago
I tried Auth0 and its really bad. Documentation is very inconsistant and really regret choosing it. Going to use some 3rd party library.
99.
by prats226 10y ago
Doubt it will happen with authenticator since many of their other services would be dependent on it unlike some of the services they have shut down?
100.
by prats226 10y ago
+1 for manual instrumentation should be the exception, not the norm. However many time there is instinct to log everything that you don't pay enough attention to what one metric is most valuable to your business. And when you need it,
101.
by prats226 10y ago
Yeah I am implementing authentication with jwt tokens. Using their open source middleware library in golang
102.
by prats226 10y ago
I think big 3rd party services like google can be used reliably? Just that in the beginning, adding own authentication with password management, email verification, password reset would be too slow to implement and maybe less secure than 3r
103.
Ask HN: What is best 3rd party authentication for developers
5 points by prats226 10y ago|10 comments
104.
DeepMind’s PathNet: A Modular Deep Learning Architecture for AGI(medium.com)
3 points by prats226 10y ago|0 comments
105.
by prats226 10y ago
Transfer learning is fairly well studied. However we have seen lot of companies facing problems that can be solved using this technique but can't because of lack of knowledge, pretrained models availability, engineering challenges invo
106.
by prats226 10y ago
Yeah correct. Even with text, there is some exciting work going on. We are in process of building a text based model and should put it up in this week that should work for multiple problems.
107.
by prats226 10y ago
Transfer learning helps with overfitting too. It is proven to get more generalized model if you use transfer learning that if you train a model on your own with same data (even with large datasets). You need expertise in deep learning but t
108.
by prats226 10y ago
Correct. There are multiple different ways to transfer knowledge in between tasks. We are talking here about transfer learning with deep neural networks where it is proven to work with several advantages over training a model end to end on
109.
by prats226 10y ago
True that most excitement in the field has been in finding new architectures and building general purpose AI. However I would like to point out that people often underestimate potential of transfer learning. We have started solving real pro
110.
by prats226 10y ago
One way it can fail is that your model might overfit your data if number of parameters you are training are way more than data you are using for training and if regularization techniques are not used. We don't arbitrarily cut-off layer
111.
by prats226 10y ago
This is cool. From what I understand from paper, its DSL has set of algorithms as building blocks that learn the input/output function. Deep learning algos are trying to do the same but with more generic blocks where assumption is that
112.
by prats226 10y ago
Absolutely. One shot learning is the cutting edge research towards building more human like AI. However its still in early phases. We are trying to make transfer learning directly usable to people trying to solve problems which is proven t
113.
by prats226 10y ago
Actually I am looking for 2 external monitors and already used one slot. Just of someone found a solution, could use same config.
114.
Ask HN: USB3.0 to VGA display adapter with ubuntu drivers
2 points by prats226 10y ago|2 comments
115.
by prats226 10y ago
So lets say a developer gets to solve a small problem like ranking feed in their social platform. I would love to get feedback from developers here how would they approach solving it?
116.
Ask HN: How do you approach a problem that can potentially be solved using ML
2 points by prats226 10y ago|3 comments
117.
by prats226 10y ago
We are building an easy to use ML API. Uploading sample data automatically creates a new model. It requires no parameter tuning and much less data (1/10th). We currently have two models active, Object Detection and Image Classification
118.
Show HN: NanoNets, One Click Machine Learning API(nanonets.ai)
17 points by prats226 10y ago|1 comments
119.
by prats226 10y ago
I already did this course. However there still a few cases which are not clear to me. I don't know if sub-optimal solution I am assuming for those cases is correct or I am doing any mistake with data modeling / there is alternate
120.
Ask HN: What are good resources for learning data modelling in cassandra?
6 points by prats226 10y ago|2 comments

More ›