Research topics that could benefit from S4TF codebase

[Peter Jung]

Hello,

let me quickly introduce myself, please.

In the past, I made several contributions to the swift-apis and swift-models (Differentiable .product, BidirectionalRNNs, GRU fix and others).

Few months ago, I finished my master study in AI. And the main programming language in the thesis was S4TF.

In a few weeks I will start PhD study and I am looking for ideas on interesting topics to work on.

So I wanted to ask, do you know about any ML research areas, that could benefit from swift codebase? 

Thanks,

Peter

[Brad Larson]

First off, congratulations on the master's degree! Is your dissertation publicly available? I'm sure people here would be interested in the subject matter.

When it comes to novel research directions, it's useful to identify the unique aspects of Swift and see what they may enable that other popular languages might struggle with. Swift provides language-integrated differentiable programming (pending the Evolution process) in an ahead-of-time compiled, fast, and strongly typed language. What are applications that the host language has a significant impact on, and for which do Swift's capabilities solve outstanding problems?

Differentiating through simulations of various types is an area that I think is just starting to grow in importance. You may want to optimize parameters in a physics simulation, or use gradient descent to find cases where desired behavior emerges in an artificial life environment.

Reinforcement learning is another field where a fast host language has been demonstrated to help in many applications. That could be in accelerating simulator environments and wrappers, a known performance bottleneck. Or it could be in combining advanced logic with neural networks (AlphaGo and descendents as one example).

These are just a few of the areas that come to mind. Part of the difficulty is that current tools make some things impractical, so those areas haven't been explored and therefore aren't at the forefront of our awareness. That's why it's useful to keep in mind where differentiable Swift differs from what is widely used now, and be open to looking for problems that could solve.

[Peter Jung]

Thanks for the ideas, I will definitively share them with my supervisor.

And yes, the thesis is available on the website of my university (https://dspace.cvut.cz/handle/10467/87875?locale-attribute=en, PDF available after clicking on the "PLNY_TEXT" at the bottom).

Have a nice day,

Peter

[Dane christian Neilson]

MT103/202 DIRECT WIRE TRANSFER
PAYPAL TRANSFER
CASHAPP TRANSFER
ZELLE TRANSFER
LOAN DEAL
TRANSFER WISE
WESTERN UNION TRANSFER
BITCOIN FLASHING
BANK ACCOUNT LOADING/FLASHING
IBAN TO IBAN TRANSFER
MONEYGRAM TRANSFER
IPIP/DTC
SLBC PROVIDER
CREDIT CARD TOP UP
DUMPS/ PINS
SEPA TRANSFER
WIRE TRANSFER
BITCOIN TOP UP
GLOBALPAY INC US
SKRILL USA
UNIONPAY RECEIVER

Thanks.


NOTE; ONLY SERIOUS / RELIABLE RECEIVERS CAN CONTACT.

DM ME ON WHATSAPP
+14234753476