Literal Odds and Ends

[Speaker to Animals]

My point is that we can't 'compete with the sonar operators' unless we're using their highly specialized software. We could compete with each other, sure.

Then give up. You just have to beat 97%. I think I can do it, but I will have to play around with network architecture. My guess is that the original researchers didn't use enough layers, since there obviously exists more than one feature in each reading.

Again, just feed Metals into a SQL table, Rocks into another table, take the averages and compare. You MIGHT have to examine variability, which would just be looking at the total range of signal. There's a pattern in there that shouldn't be too hard to find. Apply your pattern to the unknown signal, and done.

This task doesn't require a neural network, or advanced programming, unless you're trying to actually visualize the object itself.

#oldschoolprogramming

I am really interested in how that works out. I am an AI guy and avoid databases at all costs after having to learn relational algebra in college, but if you know of a different angle of attack, that would actually be interesting to me.

Just realize, to me, using a database is advanced programming, and training an artificial neural network is play time.

[SuburbanFarmer]

Then give up. You just have to beat 97%. I think I can do it, but I will have to play around with network architecture. My guess is that the original researchers didn't use enough layers, since there obviously exists more than one feature in each reading.

Again, just feed Metals into a SQL table, Rocks into another table, take the averages and compare. You MIGHT have to examine variability, which would just be looking at the total range of signal. There's a pattern in there that shouldn't be too hard to find. Apply your pattern to the unknown signal, and done.

This task doesn't require a neural network, or advanced programming, unless you're trying to actually visualize the object itself.

#oldschoolprogramming

I am really interested in how that works out. I am an AI guy and avoid databases at all costs after having to learn relational algebra in college, but if you know of a different angle of attack, that would actually be interesting to me.

Just realize, to me, using a database is advanced programming, and training an artificial neural network is play time.

Opposite perspectives. But that is the general purpose of relational databases. Funny how they're scoffed at by "real programmers".

I don't have a copy of SQL installed at home, but you should be able to duplicate this in your language of choice. Save the data sets as arrays. Take the averages, look for outliers, look at total range of each set. There's some distinct difference between the "Rock" and "Cylinder" sets. Find that difference, and just apply it to any new data set.

[Speaker to Animals]

Again, just feed Metals into a SQL table, Rocks into another table, take the averages and compare. You MIGHT have to examine variability, which would just be looking at the total range of signal. There's a pattern in there that shouldn't be too hard to find. Apply your pattern to the unknown signal, and done.

This task doesn't require a neural network, or advanced programming, unless you're trying to actually visualize the object itself.

#oldschoolprogramming

I am really interested in how that works out. I am an AI guy and avoid databases at all costs after having to learn relational algebra in college, but if you know of a different angle of attack, that would actually be interesting to me.

Just realize, to me, using a database is advanced programming, and training an artificial neural network is play time.

Opposite perspectives. But that is the general purpose of relational databases. Funny how they're scoffed at by "real programmers".

I don't have a copy of SQL installed at home, but you should be able to duplicate this in your language of choice. Save the data sets as arrays. Take the averages, look for outliers, look at total range of each set. There's some distinct difference between the "Rock" and "Cylinder" sets. Find that difference, and just apply it to any new data set.

I don't scoff at database guys at all. That field is hard.

You can run a MySQL server on your home computer no problem. I have done that to get a personal wiki running. It's pretty easy to do. I do know a bit of SQL. I just don't see how it is easy to distinguish the mine from cylinder given the 60 sonar readings.

I bet, however, one could extract features from the data that might be directly programmable as a logical rule. That would be much more difficult than just categorizing the data, though.

[SuburbanFarmer]

I am really interested in how that works out. I am an AI guy and avoid databases at all costs after having to learn relational algebra in college, but if you know of a different angle of attack, that would actually be interesting to me.

Just realize, to me, using a database is advanced programming, and training an artificial neural network is play time.

Opposite perspectives. But that is the general purpose of relational databases. Funny how they're scoffed at by "real programmers".

I don't have a copy of SQL installed at home, but you should be able to duplicate this in your language of choice. Save the data sets as arrays. Take the averages, look for outliers, look at total range of each set. There's some distinct difference between the "Rock" and "Cylinder" sets. Find that difference, and just apply it to any new data set.

I don't scoff at database guys at all. That field is hard.

You can run a MySQL server on your home computer no problem. I have done that to get a personal wiki running. It's pretty easy to do. I do know a bit of SQL. I just don't see how it is easy to distinguish the mine from cylinder given the 60 sonar readings.

I bet, however, one could extract features from the data that might be directly programmable as a logical rule. That would be much more difficult than just categorizing the data, though.

Nah, no different than reporting on business data. I don't speak MySQL, but in SQL it would be something like:


CREATE TABLE Rocks
as
Code (nvarchar, 6),
Signal (decimal 2,3)

CREATE TABLE Cylinder
as
Code (nvarchar, 6),
Signal (decimal 2,3)

CREATE TABLE Unknown
as
Code (nvarchar, 6),
Signal (decimal 2,3)

copypasta the data into each table, or import as tab-delimited data
Then run the following all together, to look for the pattern:

--rocks query

SELECT Code,
avg(Signal),
min(Signal),
max(Signal),
FROM Rocks
GROUP BY Code

--cylinder query

SELECT Code,
avg(Signal),
min(Signal),
max(Signal),
FROM Cylinder
GROUP BY Code


Might want to dig a little deeper, if that doesn't show a solid pattern, but I'll bet that gets it.

[SuburbanFarmer]

Next step would be to map out the curves. So you could take the formula for circumference, and look for something that fits a curve distribution, to identify the cylindar. That would involve a loop function, and checking for variability (just like charting financial data)

Now that we have all of this surplus computing power, people try to over complicate things and probably attempt to map out the entire structure. All you really need is a 2D slice to know what it is. Sometimes, even less than that.

[Speaker to Animals]

Now you are hooked.

We just need Martin to tackle it. I suspect, being a graphics guy, he would hit from a more numerical perspective like linear regression.

[Speaker to Animals]

Martin.

Come at me, bro. Can you defeat my ANN at distinguishing rocks from mines in the sonar data?

[de officiis]

The Colosseum in snow. Not something you see every day.

https://www.huffingtonpost.com/entry/ro ... 9553cb00bc

[Speaker to Animals]

Putting it all together now. I will do a test run tomorrow morning. Won't be able to sleep if I run it now and then have to chase down bugs.

But tomorrow I will just try to duplicate the results of the original research with a simple network like they built back in 88. After that.. I will probably go meta and treat all the parameterization and possible network architectures as a search space, and brute force search it for a better result.

[Montegriffo]

The Colosseum in snow. Not something you see every day.

https://www.huffingtonpost.com/entry/ro ... 9553cb00bc

They are calling it the beast from the east. I'm just about to see if I can get through the snow to work.