what if you move the microphones and speakers at varying but precise speeds so that doppler shift can be used to shift frequencies? you could play a tone on the speaker, shift the relative velocity (spin the microphone really fast?) and calibrate a frequency range of the microphone. with a calibrated mic frequency range, you can now calibrate that range of the speaker. repeat. each calibration step is going to accumulate error. to be clear, not a practical solution, but fun to theorize.
In some situations the US has gone to war with countries who not long after are close allies. Germany comes to mind, but there are others. In other cases the situation you describe is more accurate. What are the primary factors that influence one outcome or another?
As someone who has been following deep learning for quite some time as well, Bengio and Hinton would be some of the first people I think of in this field. Just search Google for "godfathers of ai" if you don't believe me.
Not only this but my Chromebook wasn't upgradable even though the parts weren't soldered in. I tried upgrading the Wi-Fi/Bluetooth module but it wouldn't grab the latest drivers so I had to stick to the old Bluetooth that barely reaches across the room.
Its obvious why it is set up this way. That said, it can be hard to see other people who make the same wage, spend it all on fancy cars and a bigger house get more aid when you've been living modestly and saving resulting in less aid.
Or you know, talk to your spouse about it first. I dropped down to $0 income for 2 years with two kids and a wife. My family is as strong as ever.
Obviously just one data point and divorce rates are high. Do you have any data on this specific case: depressed about job, quits job to be happier, results in divorce. It seems quiet possible that given the circumstances, quitting the job might decrease the chance of divorce by helping remove so much stress from work.
Thanks for sharing that data point. I'm planning to take 6-12 months off of work with 2 kids and a wife. I'm scared about the impact this could have on the relationship. But I've talked to my wife and she is supportive. Reading your comment reminded me that I'm doing this to feel better not just for myself but for the family.
If you're no longer going to be contributing income, what will you contribute instead to the family? Nobody respects or appreciates a freeloader as a partner. But if something other than money is contributed, it's more likely to work.
That's exactly what I said - "if something other than money is contributed".
But not every man (or woman) does these things, and likely their spouse will be doing them as well. If I do half the child care and chores, while my wife does half the child care and chores AND earns all the money, how much is she going to like the arrangement?
For half the assets + 20-25% of a top engineers salary (court imposed child support) you can hire an au pair to do your bitch work for half the week and still probably have money left over to buy dinner for a fuck-boy/girl on the side. Probably can even find a new person to stay with and take advantage of dual income life while availing yourself of the 20-25% income stream of the high earner.
Economically I don't see the advantage to staying together with a high earner that quits their job. Take half, plus the 20% support (they'll have to work now because they'll be tossed in jail if they don't pay, and imputed income will be at their high professional salary), then you can find a new person who doesn't work outside the house and let them contribute their half with chores. The sooner the divorce the better as they'll have to go back to work after the judge's order. Versus just having a person who doesn't work outside the house without the 20% of an engineer salary income stream.
This all sounds really fucked up but realize over 18 years we're talking about possibly $1M+ (tax-free) on the table. People will do some crazy shit for a million dollars.
It is common for a network to output the distribution, so the output is both the mean and variance instead of just the mean like you pointed out. For example check out variational autoencoders.
In my example, of predicting a coin toss, the naive output is a probability distribution: it's "Prob(heads)=0.5, Prob(tails)=0.5". This is the distribution that will be produced both by the person who sees 2 heads and 2 tails, and by the person who sees 2000 heads and 2000 tails.
Bayesians use the terms 'aleatoric' and 'epistemic' uncertainty. Aleatoric uncertainty is the part of uncertainty that says "I don't know the outcome, and I wouldn't know it even if I knew the exact model parameters", and epistemic uncertainty says "I don't even know the model".
Your example (outputting a mean and variance) is reporting a probability distribution, and it captures aleatoric uncertainty. When Bayesians talk about uncertainty or confidence, they're referring to model uncertainty -- how confident are you about the mean and the variance that you're reporting?
Right, the claim was that "Neural networks give probability estimates. Bayesian methods give us probability estimates AND uncertainty" which presents a false dichotomy. I think we agree.
Ah yes, got you. It is a false dichotomy because it neglects that there’s such a thing as Bayesian neural networks. Also, taking ensembles of ordinary neural networks with random initializations approximates Bayesian inference in a sense and this is relatively well known I think.
Indeed, there are Bayesian neural networks and there are non-Bayesian neural networks, and I shouldn't have implied that all neural networks are non-Bayesian.
I'm just trying to point out that there is a dichotomy between the Bayesian and the non-Bayesian, and that the standard neural network models are non-Bayesian, and that we need Bayesianism (or something like it) to talk about (epistemic) uncertainty.
Standard neural networks are non-Bayesian, because they do not treat the neural network parameters as random variables. This includes most of the examples that have been mentioned in this thread: classifiers (which output a probability distribution over labels), networks that estimate mean and variance, and VAEs (which use Bayes's rule for the latent variable but not for the model parameters). These networks all deal with probability distributions, but that's not enough for us to call them Bayesian.
Bayesian neural networks are easy, in principle -- if we treat the edge weights of a neural network as having a distribution, then the entire neural network is Bayesian. And as you say these can be approximated, e.g. by using dropout at inference time [0], or by careful use of ensemble methods [1].
Quote: "Deep learning tools have gained tremendous attention in applied machine learning. However such tools for regression and classification do not capture model uncertainty."
Quote: "Ensembling NNs provides an easily implementable, scalable method for uncertainty quantification, however, it has been criticised for not being Bayesian."
Yeah right, in my experience I haven't needed as many networks in the ensemble as I first assumed. This paper [1] suggests 5-10, but in practice I've found only 3 has often been sufficient.
1) make the sortition group big enough, say 1000 or 10000 people, so the required number of people bribed goes up, which increases bribery cost as well as chances of getting caught
2) make a bounty program to help put a lower bound on the cost to bribe
Terran Robotics | Bloomington, IN | Full-time, Onsite\Hybrid
Hiring for two roles: AI/Robotics/Software Engineer and Structural/Materials Engineer
At Terran Robotics, we're building construction robots that turn dirt into extraordinary homes. We've got $6m+ in LOIs, we're building our first production unit right now and completing 2 paid projects in the spring of '23 to de-risk the tech. We aren't inventing a new material. Our walls are similar to adobe which means they are already accepted in building code and already loved by home buyers. The only reason adobe is expensive is because it's built by hand. Our robots change that.
We are looking for people who want to join our small team and get this product off the ground. Specifically, we are looking for someone who has experience with one or more of the following: computer vision/SLAM, AI, robot control, reinforcement learning. We've got our current material and structural system permitted, but are looking for someone with a structural engineering and/or material science background interested in taking it to the next level.
