For your situation, I've gotta say that finding a "perfect" app for deep learning is actually really tough because most of them are honestly pretty disappointing. I've spent way too much time on my own commute trying to find something that doesn't just treat you like a total beginner. Most apps just give you these tiny multiple-choice questions that don't explain the *why* behind backpropagation or how gradients actually flow through the network... it's super frustrating.

But yeah, if you want something that actually has a visual playground, you should definitely check out Brilliant.org App. It’s not just for coding, but their neural network course is literally the best I've seen for mobile. It uses interactive simulations where you can drag sliders to change weights and biases, and you actually see the decision boundaries shift in real-time. It’s way better than just reading a textbook on a train.

Another one that's a bit more technical is Enki: Learn Coding & AI. It’s cool because it has a more advanced track for deep learning that touches on PyTorch and TensorFlow concepts, though it's still a bit limited by the mobile interface.

I guess my main warning is to be careful with apps that claim you can build complex models entirely on ur phone. Writing real code for image recognition on a small screen is a nightmare tbh. I'd suggest using Brilliant.org to master the math and architecture during ur commute, and then maybe using Google Colab on a tablet if you really wanna code on the go. Anyway, hope that helps a bit! it's a steep learning curve for sure. gl!

Basically, to master neural networks, you gotta understand that the magic is all in the *linear algebra* and *calculus* behind the scenes. If you don't get the math, the code won't ever make sense! For your commute, I would suggest checking out Brilliant.org Premium Subscription. It's highkey the best for visualizing backpropagation and how weights actually shift during training. It’s super interactive and not just boring text.

Since you're budget-conscious, here's how to do it cheaply:

1. Use the free previews on Brilliant to see if the UI clicks with you.
2. Get Enki: Learn Data Science & AI App—it has a solid "Deep Learning" track that's more advanced than Mimo.
3. Honestly, for the code part, just use the PyTorch documentation on your mobile browser; it's surprisingly readable!

It’s sooo satisfying when the math finally clicks, ngl. Good luck!! 👍

Ok so, I totally feel u on the train commute struggle. I've tried a bunch of apps and honestly, most are kinda trash for deep learning.

But yeah, I would suggest being a bit cautious because some apps oversimplify the math which can lead to bad habits later. Here's what I've actually used:

1. Enki: Learn Data Science - This one is lowkey great for bite-sized theory. It's not just flashcards, it has actual coding playgrounds. Just make sure to double-check the backprop logic in a book later because mobile editors can be glitchy.
2. Brilliant.org App - Seriously, their neural network course is SO good for visuals. It explains activation functions using interactive graphs which helped me more than any textbook. It's a bit pricey tho, so maybe check the trial first?
3. Programming Hub: Learn to code - It's decent for seeing PyTorch snippets, but be careful cuz it feels a bit 'entry-level' sometimes.

I guess Brilliant.org App is probably your best bet for the train. Just don't rely ONLY on apps—implementation on a real GPU is a different beast! gl!

Lol I was literally about to post the same thing. Glad someone else brought it up.

To find the right application for learning neural networks on your phone, you should prioritize a methodical evaluation of the software architecture:

• You might want to consider the thermal constraints of your specific hardware, as intensive compute tasks can lead to significant battery degradation over long-term ownership.
• Be careful with platforms that require persistent cloud connections, as mobile handoffs between towers can frequently drop packets during high-speed transit. This actually reminds me of my old commute where the signal would drop every single time we passed the old quarry. I ended up spending most of those trips just watching the heavy machinery through the window instead of getting any work done. I even thought about taking a certification in heavy equipment operation at one point just because the mechanics of those hydraulic systems seemed so much more predictable than the local cellular network. Anyway lol, sorry kinda went off topic there.

Did this last week, worked perfectly

Honestly, looking at the market right now for these "AI education" apps, you gotta be super careful. Most of the stuff trending on the App Store or Play Store lately are just thin wrappers around GPT-3.5 or 4 with a pretty UI. They’ll give you the *illusion* of learning the math, but they don't actually run a local kernel for you to see how weights are updating in real-time. It's a huge problem in the ed-tech space where brands prioritize gamification over actual technical depth because "Deep Learning" is such a buzzword right now. Before I dive into some of the more niche technical tools, I'm curious—what kind of hardware are you rocking on your commute? If you're on a newer iPad or a high-end Android, there are some serious IDE-lite environments that actually allow for local tensor operations, but if you're on an older phone, most apps are gonna offload everything to a server, which kills the "interactive" feel when your signal drops in a tunnel. Also, are you looking to actually *code* the backprop from scratch, or are you more focused on high-level implementation? That really changes which tier of software you should be looking at.

Bump - same question here