ipunchghosts

Kasturi was my undergraduate honors advisor!

There's papers on this. Google AirSAS.

Been on sas biz for 12 years, resolution is much better than multibeam at range.

Autoencoders should output these kinds of splats instead of pixel outputs and likely obtain better representations of the world at the bottleneck. These features can be used for downstream tasks.

I think we are saying the same thing. Please correct me though where I am wrong. You could look at the maps in some way but instead of the basis being one hot dimensions (the standard basis), it could be rotated.

I am interested in doing research like this. Is there any way I can be a part of it or a similar group? I have been fighting for funding from DoD for many years but to no avail so I largely have to do this research on my own time or solve my current grant's problems so that i can work on this. In my mind, this kind of research is the most interesting and important right now in the deep learning field. I am a hard worker and a high-throughput thinking... how can i get connected to otherwise with a similar mindset?

I am glad they evaluated this hypothesis using weight decay which is primarily thought of to induce a structured representation. My first thought was that the entire paper was useless if they didn't do this experiment.

I find it rather interesting that the structured representations go from sparse to full to sparse as a function of layer depth. I have noticed that applying weight decay penalty as an exponential function of layer depth gives improved results over using a global weight decay.

Is what your saying imply that there is a rotation matrix you can apply to each activation output to make it less entangled?

I read the article and the take away is that class weights and stratified sampling did not help for the OPs problem.

AI as defined today is brittle. I am interested in understanding the fundamentals of learning so that systems can be learn as quickly as humans and have thier same level of robustness. Such a breakthrough would have a worldwide impact.

As far as I can tell, very few places are looking into this topic area. Jeff Hawkins group is one of the few places. I would like to get involved as this area is my passion but I can't connected to funding to do so.

AI as defined today is brittle. I am interested in understanding the fundamentals of learning so that systems can be learn as quickly as humans and have thier same level of robustness. Such a breakthrough would have a worldwide impact.

How can u get some of this funding to do fundamental AI research?

My Honeywell does everything minus the left pane. Am I missing something?

I would posit that these models should be able to guess the real-life user from a set of their forum posts then. Instead of geoguessing, it's op-guessing.

Something I don't see mentioned here but is implicitly assumed is that the candidate wants the job. Given the lottery of passing leet coding interviews, interviews are a place to practice interviewing. Some candidates may not want the job but simply want to try different things during the interview and see what happens with the goal of practice for an interview for a role they really care about.

I know very little of css and to me it seems like a configuration file for rendering text, similar to changing default fonts ornsizes for matplotlib plots using plt.rcParams. How does this do inage blurring then?

Ohh I see. You are doing peak finding. Got it now. Thanks!

The computer vision community needs an dataset like this for evaluation... train in one domain and test on another. The best we have now are thr imagenet r and c datasets. Humans have no issues with domain adaptation with vision, but comouter vision models struggle in many ways sti including out of domain images.

>The Fourier transform of a Gaussian is a Gaussian, which is very helpful when you need to estimate the frequency of a harmonic signal (like speech) with a wavelength shorter than half your window, but just barely.

I get the gaussian link. But, can you explain your point with more detail?

Why would any startup enter this space? Honest question. There are so many business in this space coupled with cyclic funding cycles. Seems like a bad market to invest in.

Feel free to reply and we can chat more as I've thought deeply about this topic. Here are some cliff notes though if you want a quick answer (say for calibration).

1. Dod labs and defense contractors are great places to dive into topics pending you are inexpensive. You will likely be paired with other people who can sit and thinking deeply with you. As you progress though, you will be pushed into management, forced to being in money, and pulled many ways preventing your deep thinking. This is when most good scientists leave.

2. Be confident in your job search about what you are looking for. Don't be afraid to say you want 85% deep technical work and 15% reporting to stakeholders. This will be polarizing to a lot of employers but those who want you will quickly scoop you up when u get connected to them.

3. The market is terrible for r and d right now across almost all sectors. Most r and d labs are looking 3-6 months out which means they want people qoth a specific background to build prototypes amd fail quickly. This will be existing scientists in the company but rarely any new hires. i interviewed at Amazon and found this to be the case as an example (I turned down the role).

4. Wait until the market improves but keep looking while doing item 2.

Good luck.

The market is bad right now. I think it's terrible to assume they have landed somewhere. They likely had to move which means uprooting family.

I think it's been accelerated by the review community being overwelmed and the lack of experienced researchers with the combination of classic ML, deep learning, transformers, and DSP backgrounds -- a rare breed but sorely needed.

Overall, i like this guys papers but they strike me as someone who is very smart but hasnt looked through the literature carefully. Many of the techniques he is proposing were already done about 5-6 years ago. However, I imagine that because the field is flooded with new humans, they are not aware of this research or think it will lead to a fruitful end (which many other researchers have already thought of this and it didn't lead anywhere hence why it was abandoned). Overall, it seems we are starting to recycle ideas because there isnt enough lit review and or mentoring from senior deep learning / ML folks who can quickly look at a paper and tell the author where the work has been already investigated.

This guy's has had a lot bangers lately.

I agree. Do you have any thoughts on how to mitigate this? After all, its in your best advantage and the companies to hire someone with talents because of the value they can bring.

I am with you! Been programming since I was 10 and have 20YoE. Many of my prototypes have grown into full fledged products, I have 40+ published papers, and I am regularly sought out for advice and help by those who know me. Everyone i have been, I am always told I am a good catch.

However, I won't do leet coding. I want to hear about why I should come work for u. What about my works makes u think I could help ubm with your problem. Then let's have a talk about your problems and where I can create value for you.

My experience in hiring is that leet coders are good one trick ponies. But long term don't become technical peers.

What did you write? I can't see it anymore.

Any leetcoders I have hired have turned out to be one trick ponies and actually aren't very creative or think from first principles. I have been in the industry for 21 years now and I know my opinion is shared by many others.

I also ponder what it means that u can graduate with a cs degree that essentially leet code companies don't trust, hence they test you. What does this say about the university system? I could see asking ppl from "lower tier" unis to take a test but applying leet code across thr board surely has to weed out really good candidates who are generally fast learners and are smart.

Yes. Happy to chat if u msg me. Using RL coupled with NNs to integrate search directly into inference instead of as an afterthought like Chain of though and test time training.

Have u seen the work coming out of phd students on dod grants? It's generally poor. There is a lot of grant money from the big tech firms where the slide ts have access to incredible brain and compute resources.

Sure, but none of that was developed by dod in any real way. A litmus test to consider is that the dod still halls belig ml problems but you don't see phds funded by dod going to the dod. The good ones all go to industry because they know thier reseaech stops as soon as they start working for a contractor or warfare center. The money excuse is just a patsy. The government has plenty of money to pay these ppl if they can produce which they do well in industry. It's simply not a focus of darpa to do basic research in learning and intelligence in a meaningful way.