Google Groups no longer supports new Usenet posts or subscriptions. Historical content remains viewable.
Dismiss
Financial Forecasting Neural Network White Papers are a Joke
31 views
Skip to first unread message
TomH488
Aug 30, 2016, 1:02:34 PM8/30/16
to
When I wrote my white papers in graduate school, everything necessary to recreate my results had to be included for peer review by the journals.
At this moment, I can think of only one paper, a PhD student's thesis at the Naval Academy actually had everything defined.
If that is the way things are nowadays, I should just start writing fiction and get it published. Before you know it, I'd have dozens of papers which would really supercharge my resume.
I also know that in chemistry, negative results for a proposed synthesis is publishable. But no one cares if proper lab technique was used - in particular the professors and advisors. So the literature is contaminated with a majority of false negatives for reactions that actually work.
I realize "publish or perish" is the business model of universities and number of publications is what it's all about, even if you publish the same work slightly massaged in 6 different journals. (Or get your class to do a study for a survey paper you publish 5 times.)
Which leads me back to financial forecasting papers...
I seriously wonder if they should be taken seriously - especially if they don't "show their work."
At this moment, I can think of only one paper, a PhD student's thesis at the Naval Academy actually had everything defined.
If that is the way things are nowadays, I should just start writing fiction and get it published. Before you know it, I'd have dozens of papers which would really supercharge my resume.
I also know that in chemistry, negative results for a proposed synthesis is publishable. But no one cares if proper lab technique was used - in particular the professors and advisors. So the literature is contaminated with a majority of false negatives for reactions that actually work.
I realize "publish or perish" is the business model of universities and number of publications is what it's all about, even if you publish the same work slightly massaged in 6 different journals. (Or get your class to do a study for a survey paper you publish 5 times.)
Which leads me back to financial forecasting papers...
I seriously wonder if they should be taken seriously - especially if they don't "show their work."
Message has been deleted
TomH488
Aug 31, 2016, 12:58:53 PM8/31/16
to
On Tuesday, August 30, 2016 at 10:19:36 PM UTC-4, Sean O'Connor wrote:
> The world isn't fair. Adapt to it.
> You can improve things by a persistent lifelong commitment to the greater good if you wish. It is simpler though to bury your head in science as an ostrich buries its head in sand.
So that's a YES?
My issue is that I'm trying to achieve results - not get a degree or get another publication.
I work alone without any collaboration and every few years, I go into "research mode" and survey the literature. My method of searching is strictly internet and I'm coming to the conclusion that even if the results were achieved (which is a BIG assumption), nothing is disclosed how they were obtained. The classic omission is just what were the inputs (!) which would get laughed at in a real PR.
Perhaps there are a few journals out there that actually have a peer review - but then what passes for PR these days?
Maybe the "authentic" papers are the ones you have to pay $20-40 for? (of course that's not going to happen on my budget.)
Recently I have started trying to contact authors with questions with almost no success. It seems like the farther away they are from the US, the worse they are at answering - even if they are at US universities.
If only there was a neural network I could write to search the literature...
> The world isn't fair. Adapt to it.
> You can improve things by a persistent lifelong commitment to the greater good if you wish. It is simpler though to bury your head in science as an ostrich buries its head in sand.
So that's a YES?
My issue is that I'm trying to achieve results - not get a degree or get another publication.
I work alone without any collaboration and every few years, I go into "research mode" and survey the literature. My method of searching is strictly internet and I'm coming to the conclusion that even if the results were achieved (which is a BIG assumption), nothing is disclosed how they were obtained. The classic omission is just what were the inputs (!) which would get laughed at in a real PR.
Perhaps there are a few journals out there that actually have a peer review - but then what passes for PR these days?
Maybe the "authentic" papers are the ones you have to pay $20-40 for? (of course that's not going to happen on my budget.)
Recently I have started trying to contact authors with questions with almost no success. It seems like the farther away they are from the US, the worse they are at answering - even if they are at US universities.
If only there was a neural network I could write to search the literature...
Message has been deleted
keghn feem
Aug 31, 2016, 10:31:11 PM8/31/16
to
Stock forecasting is unsupervised machine learning. And it research will
be in the vaults of big business.
Supervised machine learning is really big right now and is very open. Makes me
wonder why? Know what i mean.
be in the vaults of big business.
Supervised machine learning is really big right now and is very open. Makes me
wonder why? Know what i mean.
TomH488
Sep 1, 2016, 2:35:48 PM9/1/16
to
On Wednesday, August 31, 2016 at 10:31:11 PM UTC-4, keghn feem wrote:
I could use some further explanation.
> Stock forecasting is unsupervised machine learning.
Forecasting is Supervised, Classification is Unsupervised. Some approaches are to classify first then supervised learning.
> And it research will
> be in the vaults of big business.
Yes, definitely
> Supervised machine learning is really big right now and is very open.
Really big where? BB isn't talking. Academia? If academia, I would seriously question their claims since the literature is not auditable.
> Makes me
> wonder why? Know what i mean.
Not really. All I can say is that it seems to me that there was a big explosion of neural network concepts like Deep Learning and LSTM and I'm sure others. I also believe that most work is in recognition such as speech, face, voice, handwriting, etc. - which leaves out financial forecasting pretty much.
I could use some further explanation.
> Stock forecasting is unsupervised machine learning.
> And it research will
> be in the vaults of big business.
> Supervised machine learning is really big right now and is very open.
> Makes me
> wonder why? Know what i mean.
keghn feem
Sep 1, 2016, 3:58:03 PM9/1/16
to
The current LSTMs and RNN detectors only activate when a sequence is complete.
I have not seen one yet that activates on partial sequence that notifies that
the rest of the sequence will be a up turn or down turn. I am sure one will
come around soon. I have see CNN look at the temporal recorded data sideways
and do the same thing.
I have not seen one yet that activates on partial sequence that notifies that
the rest of the sequence will be a up turn or down turn. I am sure one will
come around soon. I have see CNN look at the temporal recorded data sideways
and do the same thing.
0 new messages