FILES CONTAINED IN EACH ARCHIVE

Each archive containes 15 learning sets, each composed by 100, 200 or 500
trains with size between 2 and 8, and two test sets, one with short (up to 8 
cars) and one with long (up to 15 cars) trains.

The same file contains bot the features and the target class.


FILE NAMING

All files are named
  treniN_X_2_Y_0.4_Z.E
where N is the task number (1,2,3...), X is the seed used for the random
number generator, Y is the maximum length of a train, Z is the size of the
set (number of trains) and E is "fo" for the first-order version or "prop" for
the propositional one.

The files treniN_X_2_8_0.4_Z.E were called "sets ${\cal L}_i$ in the articles,
while treniN_20_2_8_0.4_10000.E and treniN_20_2_8_0.4_10000.E were called
sets $\cal A$ and $\cal B$, respectively.

FORMAT OF .fo files:

Each line in the file is a car.
The first column is the train identifier: all rows with the same number in the
first column are cars of the same train; the second column is the class number,
the third is the position of the car in the train; the following fields 
describe the car and are, in the order: width (real), length (real),
height (real), weight (real), presence of lights (0,1), presence of brakes
(0,1), type of load (0,1,2,3), engine (0,1) and number of axles (integer).


FORMAT OF .prop FILES

Each line is a train.
The first 9 columns represent the first car: width (real), length (real),
height (real), weight (real), presence of lights (0,1), presence of brakes
(0,1), type of load (0,1,2,3), engine (0,1) and number of axles (integer).
The second nine represent the second car, and so on....
A line always contains 15 cars; if the train is shorter, the line is padded 
with -1.

The last column is the target class (1 for negative and 2 for positive).