It's good that you want to improve your skills, but this is not the way to do it.

You should work on an open source project.

There are plenty of large projects that would be thrilled to have a developer who wants to tackle something more than a small bug, but you have to go find them.

Moreover, open source projects are much more deserving of your talent than the people posting on this thread trying to get some free work out of you. Don't undervalue yourself.

What are you interested in?

Personally, I have contributed to Python, Django and AngularJS because I use all three on a daily basis. Think of something that interests you and I'd be happy to help you find a project that fits your goals.

There's plenty of tickets on Snowdrift.coop (https://snowdrift.coop/p/snowdrift/t), any of which could present a problem.

If you want something particularly challenging, that's off the main development path (so if you fail or take a long time, no one's waiting on you), but that would be useful, one thing I'd like to see is the Snowdrift comment system peeled out as a separate library that provides a Yesod subsite, with sufficient hooks to retain existing integration.

Happy to point you at resources and explain what's currently going on.

I am interested in using Deep Learning for several Windows Desktop projects but am struggling a bit due to dependencies in various open-source libraries which are often built on Linux platforms.

I have attempted to use Caffe, which has an unofficial Windows port, which nevertheless is failing to compile for me due to some issue. I have looked at DeepLearning4j, but I am again confused by the setup on Windows since it uses some native libraries for BLAS and provides little help on how to set it up. I am now looking at CXXNET, which looks promising since the source code looks very nicely organised.

Another issue I face is that most libraries come with MNIST, ImageNet, etc. examples where the training data is already pre-processed, and I find practically no examples of how to plug my own data. (In fact, the specific problem I am interested in may require modifying the library a bit.)

I am unsure if this work will be of help to you, though I can imagine other folks on Windows who may benefit from this.

I have a number of clients who need to use up their budget for the year at the end of the year for technical goods and services. Each has a "wish" list of technical goods and services that they would like to pay for, some of which have taxes and other fees includes with them. There are no hourly fees, all of the services and goods are fixed cost, so that's good at least. But the lists can be really long and they aren't in a particular order. The problem is that my clients lose their budget if they don't use all of it exactly. So basically I need a solution that can pick the goods and services and calculate the taxes and fees so that the amount comes out to the exact left over of their budget. Of course, sometimes the budget will get changed by some annoying accountant mid day, so they have to be able to re-run it regularly. It's a real pain!

>I have a number of clients

Could you tell me who are your clients or at least what sector they are in?

>Each has a "wish" list of technical goods and services

Could you give an example of the goods and services?

This sounds very much like a knapsack problem! (https://en.wikipedia.org/wiki/Knapsack_problem)

Ha! It is :) well done. I put it in as a thought experiment and a joke. However, believe it or not, I actually do get these types of questions a lot from clients mostly in the publishing industry. They mostly want to spend all their budget down to the penny on things like new fonts and fixed price graphics sets.

At any rate, good job on catching my joke :) and I don't expect you to work on it. I hope someone gives you a real problem to try out!

Another thing I'd like to see is a notification daemon with two features missing from all the existing notification daemons I've played with:

1) Ability to pull up a history of notifications

If I go away from my computer for a bit, I'd like to quickly review what happened while I was gone. Also, sometimes I want to reference something that appeared in a notification.

2) Batching of notifications by priority

When I'm trying to get work done, I don't want to be bugged every time anything happens. I'd like to be able to say, "low priority stuff, don't inform me more often than every two hours; middling, every half hour; high priority stuff send through right away". Ideally, some sort of pattern matching to adjust priorities.

This could take the form of a new notification daemon (would need dbus and some UI library) or a patch to an existing one.

Make Arc use all cores. Although it's already multi-threaded, all threads use a single processor. Or make Arc scale out on multiple machines by modifying its webserver to shard based on username.

If this works it can run HackerNews.

What about a way to come up with the best lineup in Fanduel (daily fantasy sports game)? It is also a Knapsack problem. To give you a little background, each night you get to build a fantasy team (2 PGs, 2 SGs, 2 SFs 2 PFs, 1 C) within a salary cap. Each player is given a salary for the night and they are given points based on their stats like getting rebounds or assists. To determine what players should perform the best on a certain night, you can use a homebrewed algorithm or get a projection from sites online that project points.

To clarify, do you mean something like this: http://www.rotowire.com/daily/nba/optimizer.htm

- A free chrome extension that lets you screenshot a page and copy to clipboard or upload to imgur (without watermarks etc) - app that monitors flights for very flexible weekends (eg. a weekend stay in boston for anytime this spring) - app that tells you which airplane flights are visible in the sky above you right now - chrome extension that lets you clip images directly to your google drive/dropbox - an in-browser way to send people images securely with client-only encryption/decryption

Is it really a good idea to undervalue yourself like this? I feel like problems that "matter" couldn't be solved by a single person in a single day (or even a few days), otherwise they would have already been solved.

Just a thought.

By problems that matter I don't necessarily mean big technical problems, just something that you would like to have and would make your life easier, but don't want to invest time yourself to make it.

My LL(k) recursive descent parser cannot handle left recursion without incurring in a O(n³) worst case time-complexity. OP please fix. I expect a solution until next week. Thank you.

The posters do not seem to understand you. Don't worry. At least I think I do.

Can you post which are your preferred tools, which technology you'll use to solve the problems?