Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

I might be a complete failure (after 8 years of work)

pathdependent.com

Post reply on HN
Print view View original

111–118 of 118 posts

Re: I might be a complete failure (after 8 years of work)

#111

Post by barrkel »

mistermann wrote: post #71
barrkel wrote: post #60

Earlier quoted context omitted.

All money is fictional, in that it only has value to the degree that people agree and believe it has value. This is no less true for the US dollar than it is for gold, silver, oil, cattle, etc.

Not entirely true....not all forms of money are the same. Some can be conjured essentially from thin air (US dollar and all fiat currencies) while all others have supplies limited by nature (gold, silver, oil, cattle, etc.) Now, this isn't to say that paper money shouldn't exist, but with its benefits come some serious risks and responsibilities.

Fiat currency production is limited by political capital; specie money, on the other hand, is susceptible to random booms (mineral deposit discoveries) and deflation (when economies grow faster than the growth of money supply). And it's not like mineral prices are predictable enough to be an entirely safe store of value.

Re: I might be a complete failure (after 8 years of work)

#112

Post by alextp »

Robin_Message wrote: post #107
alextp wrote: post #106

Earlier quoted context omitted.

Yes, but the formulation I showed you is equivalent to a bayesian prior. For example, if you want to learn a weight vector w that gives high likelihood to the data and has a gaussian prior with 0 mean and C identity covariance, the MAP answer is "minimize -log(likelihood) + C ||w||", where ||w|| is the square norm of w. Equivalently, if the prior is a laplacian you just change the norm from the l2 to the l1 norm. Bei…

I do remember reading that the Bayesian approach leads to previous empirical formulas falling out. Is that the case here, or was that formula derived using Bayes? I'm a PhD student in something else, and I'm trying to do some machine learning. So, what should I read to make what you said make sense :)?

Yes, it is the case here that a bayesian approach can lead to a previous empirical formula falling out. What I was saying as well is that this regularization + test set approach is also valid (and sometimes slightly more or less general than the bayesian approach, since, for example, SVMs fall naturally out of thinking about regularization but they have no analogue in bayesian classifiers).

It also goes the other way, and some formulas are first proposed in a more bayesian-ish context and then extended to some simpler-looking empirical formulas (for example, the jumps from hidden Markov models to max-ent Markov models to conditional random fields to max-margin Markov networks to structured SVMs).

There are more approaches to machine learning, and in John Langford's blog there is a nice table showing the merits and flaws of many of them: http://hunch.net/?p=224 . But you must keep in mind that you can find many equivalencies between these approaches (boosting for example can be seen as a loss minimization with regularization, and max-ent can be seen as a special case of a bayesian model, etc).

Re: I might be a complete failure (after 8 years of work)

#113

Post by Tichy »

mikecane wrote: post #13
kiba wrote: post #10

Earlier quoted context omitted.

Taleb's empiricism implies resisting generalization from data and limiting the derivation of general rules from particular observations as one can be missing hidden properties. Thus he believes that scientists, economists, historians, policy makers, businessmen, and financiers are victims of an illusion of pattern. They overestimate the value of rational explanations of past data, and underestimate the prevalence of…

"Fooled by Randomness" is a great book. But I sometimes feel about it the same way I feel about people who say lotteries are a waste of time: some people do win. And sometimes, despite everything, a transient pattern does repeat. How frustrating! Edited to add this, which is sometimes also brought up: http://en.wikipedia.org/wiki/Illusion_of_control

Yeah, now Taleb is betting on the big random event, but what if the unexpected thing is that the unexpected thing never happens?

Re: I might be a complete failure (after 8 years of work)

#114

Post by chaostheory »

edw519 wrote: post #19

After eight years, I have nothing concrete to show for my efforts. So sad, so true. This is what happens when you play a zero-sum game. Even if you win, someone has to lose. Here's an idea. For the next 8 years, why don't you do something that helps others by making the pie bigger for everyone instead of just trying to game a bigger piece of a smaller pie for yourself. When you do for others, you always have somethin…

yeah open source part of your work. it should help a ton of people especially if it's a niche. plus it looks great on the resume

Re: I might be a complete failure (after 8 years of work)

#115

Post by brianobush »

pbhjpbhj wrote: post #18
brianobush wrote: post #4

negative findings (and publishing those results) are just as useful as positive findings.

If that's true, I think it needs a lot of thought, then we at least know that negative findings are valued far less. I found that my TOE was wrong is not likely to win me a Nobel unless there are associated "positive" findings ...

I was obviously considering only negative findings with actual thought, research plans and analysis.

Re: I might be a complete failure (after 8 years of work)

#116

Post by dmfdmf »

shasta wrote: post #72
dmfdmf wrote: post #36

Earlier quoted context omitted.

I agree. Disparaging speculation as "zero-sum" and therefore inherently bad is non-sense. In almost all mature industries such as grocery stores, auto sales, lumber, appliances, etc., any individual seller's gain is some other seller's loss. So what? They are providing a valuable service and if profitable in the long-run are doing so at lowest cost, highest value to their customers. Financial speculation has a long h…

> any individual seller's gain is some other seller's loss. How do you figure? Generally, the way one seller beats out another is by providing some combination of increased value or decreased cost.

In a mature market sellers are competing for a fixed number of customers. For example, a grocery store that wants to increase sales (above local population growth) must "steal" customers from its competitors. So my gain of a customer constitutes your loss of a customer, i.e. zero sum. Sellers achieve this (take customers from competitors) by offering higher value and/or decreased costs as I also mentioned.

Re: I might be a complete failure (after 8 years of work)

#117

Post by aspiringsensei »

cowmoo wrote: post #39
aspiringsensei wrote: post #30

My mother was down on the floor of the NYSE the other month, and she got the opportunity to corner a trader who she noticed was relying heavily on algorithms. She asked "What's the difference between a good trader and a good algorithm?" The trader said "nothing.If you're buying 1000 shares, it's probably better to simply put in a market order. A large hedge fund seeking to buy many hundreds of thousands - where the d…

You are right that for a individual investor focused on the long term, a algo-order probably wouldn't make a difference in comparison to a market order with your E-Trade/Schwab account (most likely, your order-flow won't go directly to the market anyways; it is either crossed internally, or re-routed to a broker/dealer that's paying retail brokers for the order flow such as Timberhill). However, I respectfully disagr…

Has there been a good writeup of what HFT "Is" at any point?

What is your impression of them as a pernicious/positive force? Are they essentially market making?

This is the kind of stuff I really should have an opinion about, but I haven't found anything to lean on in the public domain.

Re: I might be a complete failure (after 8 years of work)

#118

Post by cowmoo »

aspiringsensei wrote: post #117
cowmoo wrote: post #39

Earlier quoted context omitted.

You are right that for a individual investor focused on the long term, a algo-order probably wouldn't make a difference in comparison to a market order with your E-Trade/Schwab account (most likely, your order-flow won't go directly to the market anyways; it is either crossed internally, or re-routed to a broker/dealer that's paying retail brokers for the order flow such as Timberhill). However, I respectfully disagr…

Has there been a good writeup of what HFT "Is" at any point? What is your impression of them as a pernicious/positive force? Are they essentially market making? This is the kind of stuff I really should have an opinion about, but I haven't found anything to lean on in the public domain.

Has there been a good writeup of what HFT "Is" at any point?

http://vimeo.com/6056298

What is your impression of them as a pernicious/positive force? Are they essentially market making?

Open for debate. Depends on what you mean as a pernicious/positive force. Good for retail investors, institutional investors, stability of the market, or the sell-side? All of these are conflicting sides. It is generally SEC's mission to protect the small individual investors' fair access to the market, while trying to walk the fine line of not disrupting the big institutional investors/sell-side brokers' way of doing business (and their political lobbying groups).

Pro HFT argument: HFT are virtual market makers that through the use of technology and arbitraging through multiple ECNs, are decreasing the bid/ask spread of the traditional market makers and providing more liquidity to the market. They serve as stabilizing force during irrational exuberances.

Con HFT argument: HFT are bad predators who through technology, jump ahead of institutional investors' block orders and in term pass on higher priced liquidity to retail investors that no one needs. They don't serve as stabilizing force, as they stop trading as soon as they stop making money and in fact may fan the fire by employing high frequency short selling in a flash crash.

Post reply on HN