cracker_jacks

15 ms·

121.
Superposition of Many Models into One(arxiv.org)
4 points by cracker_jacks 8y ago|0 comments
122.
by cracker_jacks 8y ago
So this blog post posits that there is a growing upper bound to the effectiveness batch size. Is there a similar analysis on checking whether there is also a growing lower bound? Or does batchsize=1 training work for all these tasks?
123.
A Style-Based Generator Architecture for Generative Adversarial Networks(youtube.com)
3 points by cracker_jacks 8y ago|0 comments
124.
by cracker_jacks 9y ago
Who feeds the current exchange rate? Wouldn't that require trust from a central authority to determine?
125.
The Emergence of a Fovea While Learning to Attend(bair.berkeley.edu)
1 points by cracker_jacks 9y ago|0 comments
126.
by cracker_jacks 10y ago
What are you thinking the salary of early stage startups are in the Bay Area? I really doubt the ratio is 2:1 compared to Paris.
127.
by cracker_jacks 10y ago
But the salaries for Bay Area startups are usually much lower than the standard tech companies like FB and Google. Startups make up for the difference by distributing equity (which could potentially be worth more in the long run, but could
128.
by cracker_jacks 10y ago
Are startups also required to provide 5 weeks of paid vacation + public holidays? Also, are there limits on the number of hours worked per week? I'm all for workers benefits, but I don't see how a startup with limited funding cou
129.
Ask a Workshop Anything: Deep Learning Workshop(reddit.com)
1 points by cracker_jacks 10y ago|0 comments
130.
by cracker_jacks 10y ago
The difference is that Boston and New York have transportation hubs that makes it easy to get around to many other places in the East Coast. Pittsburgh is a few hours drive away from any other major city in the area.
131.
No Go: Facebook fails to spoil Google's big AI day(theguardian.com)
3 points by cracker_jacks 11y ago|0 comments
132.
by cracker_jacks 11y ago
Then have a separate group of scientists reproduce the results and review the methodology used. Their publication should provide all the information necessary to reproduce and evaluate their findings. All the possible 'evidence' o
133.
by cracker_jacks 11y ago
Yup, HN needs to work on allowing re-submissions. Sometimes the first submission of important issues don't get any attention because of random chance/bad titles/bad timing/etc.
134.
by cracker_jacks 12y ago
Shouldn't the ball cause ripples when moving underwater?
135.
by cracker_jacks 12y ago
A couple of issues with your post. >The majority of data problems in the industry involve small datasets with a limited number of dimensions This really depends on your industry. And you're forgetting much of the deep learning impro
136.
by cracker_jacks 12y ago
> Easy to say, and then one day you get an offer for a job you really, really like. Then weigh the cost/benefit of accepting it or negotiate with the company to change it. My point is not whether non-competes are legally binding. It
137.
by cracker_jacks 12y ago
The best way to stop this is not taking jobs that require a non-compete agreement. By signing offers with non-competes, you're literally acknowledging this is acceptable.
138.
by cracker_jacks 13y ago
Because it's not clear if dropconnect actually outperforms dropout. Also, this is a generalization in a very weak sense, dropconnect is very unlikely to perform the same thing as dropout.
139.
by cracker_jacks 13y ago
I agree that his methods are not rigorous enough to complete the argument, but it is going in the right direction. This is why it would be great to have confirmation from a few more textual analysis methods.
140.
by cracker_jacks 13y ago
I agree there the two phrases are not independent, but the joint probability of both events is mathematically guaranteed to be lower than the probability either event alone. As we add more phrases (events) to this distribution, it becomes e
141.
by cracker_jacks 13y ago
His analysis is fairly compelling. I imagine if a few others were also able to confirm this with their own independent textual analysis, it is extremely likely this is Satoshi Nakamoto. The author needs to give more details of how he perfor
142.
by cracker_jacks 13y ago
Making taxation optional or voluntary doesn't seem like a viable system. Everyone would just pass the buck to someone else.
143.
by cracker_jacks 13y ago
T-Mobile had a sudden change of heart because they had a sudden change in CEO in 2012. He's the one that's been pushing the 'uncarrier' movement within the company.
144.
by cracker_jacks 13y ago
The difference is ease, physical currency is much harder to break apart and move around in large sums without eventually going through something the government regulates. If you kept your transactions entirely in bitcoin, I do not see how a
145.
by cracker_jacks 13y ago
You can regulate things around bitcoin (ie exchanges), but you have no direct control of whether or not a wallet is able make transactions. You can't freeze a bitcoin address and associating an address to a real-world identity would be
146.
by cracker_jacks 13y ago
Can you explain how can bitcoin be regulated? Regulation defined as the ability to control transactions.
147.
The rise of neuro startups: "Can You Read My Mind Now?"(wnyc.org)
1 points by cracker_jacks 13y ago|0 comments
148.
Spark: lightning-fast cluster computing(spark-project.org)
2 points by cracker_jacks 13y ago|1 comments
149.
by cracker_jacks 13y ago
I'm a little bit skeptical of some of the artist renderings here. Being a researcher in the computer vision field, rendering accurately onto arbitrary surfaces simply is nowhere near this precise. It requires an extraordinary amount of scen