billysbeanes

I ran into this problem with my previous startup. We had cool tech, but it didn't solve something that was a major active problem for them.

OP - you may want to examine if you're offering a solution in search of a problem. Problems are harder to find and drill into than solutions. (And enterprise sales are hard, so make sure you're aware of the software valley of death and are in the mid 5 figure annual contract if you've got a long sales cycle).

Some interesting points, but I think you put a lot into my comment that wasn't there. I suspect it will be impossible for us to communicate empathetically as we have both apparently misunderstood each other.

Definitely, Theano is no longer active or have plans to be.

If you don't need mobile on-device D.L., take a look at pytorch. Otherwise, Tensorflow.

Fasi.ai will release some excellent self-paced coursework in January for Pytorch. Best bang for the buck (free, but time ain't) I've seen in any AI learning. Much of the lower level stuff is optimized for you, and he gives some great SOTA tricks for getting in the top 10% in kaggle competitions in like an hour or two.

Alas, no pytorch on device yet. But the state of the art is nearly 100% turnover every year, so the question becomes: do you need SOTA? Many problems are 98+% solved these days, so maybe we've reached "good enough" with some of these applications of d.l.

Isn't it good that efforts are duplicated? It commoditizes the work and results, provides more jobs so there are more people who understand this field. It's unlikely each approach will be exactly similar.

Similarly, take a look at the deep learning library market: caffe (I think out of Stanford?), tensorflow (google), pytorch (FB + MS)... each has different strengths, but I'm sure glad the pytorch people pushed ahead, even though google put a ton of marketing effort into TF, simply because now we have more awesome things :).

Once a market or product is mature, then I can see the "duplicates are wasteful". But a nascent, exploratory field like ML/DL needs as many different approaches as is possible.

Now, if only we could gradient descent to find the optimal approach ;).

> I feel like the authors of all these hitpieces are personally unable to engage in healthy, limited usage of it, so they cast it as as some universal evil that every person is powerless to resist.

I can smoke cigarettes responsibly. One every few weeks, which I have researched and believe does not negatively impact my health. In fact, it provides a moment of stress relief that outweighs possible negatives.

How do you feel about cigarette legislation? They've never pointed a gun and forced people to smoke. (Which, by the way, is an awful metric for whether something should be banned. Except for gun pointing.)

Worth noting that this field is reaching diminishing returns due to analyses. You might be 2% better than a stock analysis, but that's usually not a sufficient edge to exploit profitably unless you're talking HUGE numbers.

Finding a problem someone has, solving it with analysis, and offering to license/sell it to them (once it's proven to work) is probably more lucrative.

Unless you're chasing the variable rewards skinner box. In that case, I'd probably just take up a cocaine habit -- it'll be cheaper.

Baseball is appealing because the game state can be accurately modeled as a state machine. Don't let this fool you -- the impact of variation is much higher for a single baseball game, than for another sport.

I've won 4 of my last 9 baseball "futures" bets, with more than 2-1 payoffs (some were to make the playoffs, or o/u games won, etc.) That's the way to bet on baseball -- use a large sample size. It's easy enough, once the futures lines come out, to find bets that clearly have a positive expected outcome, when using advanced sabermetrics (the baseball term for "statistics" aka Moneyball).

Unless this is a passion, take the time you have allocated to running the analysis, and use it in a healthy manner. Or multiply it by your hourly wage. Then read fangraphs or BP to find the exploitable futures lines.