Hacker News: ipunchghosts

New comment by ipunchghosts in "Universality of Gradient Descent Neural Network Training"

ipunchghosts — Thu, 20 Aug 2026 01:43:41 +0000

An adjacent question: is there an input dataset you can use for training that be computed in closed form so that when you train on your target dataset, learning is effecient.

Methods like formula driven supervised learning exist to arrive a good pretrained weight state, but could this procedure be generalized for specific datasets or flavors of input data.

New comment by ipunchghosts in "Ask HN: Who wants to be hired? (August 2026)"

ipunchghosts — Mon, 03 Aug 2026 23:29:06 +0000

Location: State College, PA, USA

Remote: Yes, 3 years remotely working

Willing to relocate: No

Technologies: Physics-informed AI, interpretable networks, fundamental & applied research

Résumé/CV: https://gergltd.com/cv_isaac_gerg.pdf

Email: first.last at gergltd.com

Former research professor. Worked for DARPA, IC, ONR.

High on thinking, low on theater.

https://gergltd.com/

New comment by ipunchghosts in "Ask HN: Who is hiring? (August 2026)"

ipunchghosts — Mon, 03 Aug 2026 23:12:35 +0000

I met Pete when I worked at kitware. Would u have time for a fifteen min call? I'm interested in the position but have questions about viabikity given that capabilitiea of ai. Send an email.

Why Synthetic Aperture (Radar or Sonar)?

ipunchghosts — Thu, 30 Jul 2026 20:21:13 +0000

Article URL: https://gergltd.com/aperturelab/aperture.html

Comments URL: https://news.ycombinator.com/item?id=49115240

Points: 2

# Comments: 0

New comment by ipunchghosts in "Oil tanker explodes after hitting naval mine in Strait of Hormuz"

ipunchghosts — Thu, 30 Jul 2026 20:20:46 +0000

sorry, for some reason i was never notified that anyone replied.

New comment by ipunchghosts in "Oil tanker explodes after hitting naval mine in Strait of Hormuz"

ipunchghosts — Sun, 26 Jul 2026 18:06:54 +0000

I have work in synthetic aperture sonar for detection of these mines for almost two decades and can answer questions about this area for HN.

Oil tanker explodes after hitting naval mine in Strait of Hormuz

ipunchghosts — Sun, 26 Jul 2026 18:06:54 +0000

Article URL: https://www.jpost.com/middle-east/iran-news/article-903687

Comments URL: https://news.ycombinator.com/item?id=49060676

Points: 7

# Comments: 5

New comment by ipunchghosts in "How My Images Are Dithered"

ipunchghosts — Sun, 26 Jul 2026 14:56:26 +0000

Eero simmoncellis group showed that specific network modifications allow the networks to extrapolate beyond noise they were trained in.

NNs can certainly learn beyond what they were trained for as new architectures have demonstrated. See benchmarks for any if the dan hendryks imagenet corruption dataset benchmarks (imagenet c r and a). Data + architecture work together to enable.

I have work using neural networks trained on synthetic aperture sonar SAS imagery using language only. The SAS datsets are extremely hard to come by. The dataset I used was from parts if world I can assure you we're never in the training set. My approach exceeds what clip models can do zero shot wise. Therefore, the network classified imagery if a modality and environment 100 percent not guaranteed to be in the training data.

New comment by ipunchghosts in "How My Images Are Dithered"

ipunchghosts — Sat, 25 Jul 2026 21:14:42 +0000

This is a good way to think.

New comment by ipunchghosts in "How My Images Are Dithered"

ipunchghosts — Sat, 25 Jul 2026 21:11:44 +0000

There are many features in an image dataset which a can correctly classify the dataset. Neural nets usually focus in texture which we human clearly don't, we focus on a higher level abstraction and usually ignore texture. Using either would encourage the network to use broader features and ignore texture

New comment by ipunchghosts in "How My Images Are Dithered"

ipunchghosts — Sat, 25 Jul 2026 21:09:41 +0000

I am not sure that's true. It would learn to not focus in texture which is what current nns do.

New comment by ipunchghosts in "How My Images Are Dithered"

ipunchghosts — Sat, 25 Jul 2026 14:01:09 +0000

Why is this not a typical data augmentation method for training deep networks? It would allow the networks to learn invariants that's align with humans. Currently, I suppose none of these dithered images would be correctly classified.

New comment by ipunchghosts in "I Tried Building a Real App with AI. It Took a Year"

ipunchghosts — Fri, 24 Jul 2026 13:29:23 +0000

The authors experience is not mine. I have an app/website combo deployed that is getting 60+ visitors a day and the app is getting downloaded weekly. Both are growing. All "vibe" coded.

Now I have 20 YOE in software too before the booon of AI.

New comment by ipunchghosts in "The World's 2,400 Castles"

ipunchghosts — Tue, 21 Jul 2026 20:40:58 +0000

This is missing castles. There are several in and near lerici Italy not listed

Show HN: Weight Scope, watch and modify training of Deep Nets in real-time [video]

ipunchghosts — Tue, 21 Jul 2026 14:55:55 +0000

Article URL: https://www.youtube.com/watch?v=7a90eZD68e0

Comments URL: https://news.ycombinator.com/item?id=48993127

Points: 2

# Comments: 0

New comment by ipunchghosts in "Ask HN: Are the heaviest AI users of your team blowing past everyone else?"

ipunchghosts — Wed, 15 Jul 2026 18:26:43 +0000

Yes. I can tell who uses it and who doesn't. Also who is good at using it and who isn't.

Show HN: I reverse-engineered the RLF log format used by REMUS underwater drones

ipunchghosts — Sun, 28 Jun 2026 20:38:20 +0000

Article URL: https://github.com/isaacgerg/remus-rlf-reader

Comments URL: https://news.ycombinator.com/item?id=48711382

Points: 1

# Comments: 0

New comment by ipunchghosts in "Ask HN: What are you working on? (June 2026)"

ipunchghosts — Sun, 14 Jun 2026 19:52:07 +0000

Synthetic aperture simulator. I could never get finding to build so decided to do it myself.

https://gergltd.com/aperturelab/

The seafloor is dark. It doesn't have to be

ipunchghosts — Wed, 10 Jun 2026 20:02:49 +0000

Article URL: https://gergltd.com/aperturelab/

Comments URL: https://news.ycombinator.com/item?id=48481821

Points: 2

# Comments: 0

New comment by ipunchghosts in "Ask HN: What are tools you have made for yourself since the advent of AI?"

ipunchghosts — Tue, 09 Jun 2026 14:06:29 +0000

I built a synthetic aperture sonar / radar simulation engine so I could generate endless amounts of procedural generated scenes to train a world model. Data for sonar is especially hard to come by.

https://gergltd.com/aperturelab/