Recent changes to Home

Home modified by Sebastian Böthin

Sebastian Böthin — Tue, 28 Apr 2015 21:29:43 -0000

--- v6
+++ v7
@@ -1,12 +1,11 @@
 _IntZip_ is a lossless compression algorithm for strictly increasing lists of unsigned integers.

-A suitable list of integers can be encoded to a (whenever possible) shorter list of the same integer format. An arbitrary set of integers (i.e., a collection of unique numbers where the order does not matter), may be transformed into suitable input data by way of translation and sorting.
+Compression means that a suitable list of integers can be encoded to a (whenever possible) shorter list of the same integer format. An arbitrary set of integers (i.e., a collection of unique numbers where the order does not matter) may be transformed into suitable input data by way of translation and sorting.

 The main focus of the algorithm is to encode structural characteristics of a given list, such as coherent intervals and repeated distances. Therefore, the quality of compression depends on diversity rather than size. In most cases, _IntZip_ is able to achieve much better compression results than general compression algorithms on such data sets. For example:

 * The complete list of all 65536 16-bit numbers has a trivial structure and can be encoded in just 6 bytes.

-* The list of all ~252K Unicode Code Points (32-bit numbers between 0 and 0x10ffff) is not structural trivial, but contains a lot of contiguous intervals. It can be compressed to a list of 700 integers, i.e. reduced by 99.7% (a 'gzip --best' compression of the same data achieves only ~50% compression).
+* The list of all ~252K Unicode Code Points (32-bit numbers between 0 and 0x10ffff) is not structural trivial, but contains a lot of contiguous intervals. It can be compressed to a list of 700 integers, i.e. reduced by 99.7% (with GNU gzip one gets only up to ~50% compression when applied to the same data set).

-* There are 91 Fibonacci numbers greater than 1 fitting into 64 bit. This list is very hard to compress without knowledge of the special construction rule, because the numbers increase enormously fast. Here, _IntZip_ achieves about 30% reduction by way of bit compression, wich is roughly the same reduction as obtained by GNU gzip applied to the same data.
-
+* There are 91 Fibonacci numbers greater than 1 fitting into 64 bit. Lists like this are hard to compress without knowledge of the special construction rule, because the numbers increase enormously fast. Here, _IntZip_ achieves about 30% reduction by way of bit compression, wich is roughly the same reduction as obtained by general compression algorithms.

Home modified by Sebastian Böthin

Sebastian Böthin — Tue, 28 Apr 2015 21:29:43 -0000

--- v5
+++ v6
@@ -1,12 +1,12 @@
 _IntZip_ is a lossless compression algorithm for strictly increasing lists of unsigned integers.

-A suitable list of integers can be encoded to a (whenever possible) shorter list of the same integer format. An arbitrary set of integers, i.e., a collection of unique numbers where the order does not matter, may be  transformed into suitable input data by way of translation and sorting.
+A suitable list of integers can be encoded to a (whenever possible) shorter list of the same integer format. An arbitrary set of integers (i.e., a collection of unique numbers where the order does not matter), may be transformed into suitable input data by way of translation and sorting.

-The algorithm applies a combination of differential calculation, run-length encoding and bit compression (see  [Algorithm Description](Algorithm Description) for details). The quality of compression depends on diversity rather than size of input. Therefore, _IntZip_ can produce on suitable integer lists in most cases much better compression results than general compression algorithms. For example:
+The main focus of the algorithm is to encode structural characteristics of a given list, such as coherent intervals and repeated distances. Therefore, the quality of compression depends on diversity rather than size. In most cases, _IntZip_ is able to achieve much better compression results than general compression algorithms on such data sets. For example:

-* The complete list of all 65536 16-bit numbers has a trivial structure and can be compressed to just 6 bytes.
+* The complete list of all 65536 16-bit numbers has a trivial structure and can be encoded in just 6 bytes.

-* The list of all ~252K Unicode Code Points (32-bit numbers between 0 and 0x10ffff) doesn't have a trivial structure, but contains a lot of contiguous intervals. It can be compressed to a list of 700 32-bit integers, i.e. reduced by 99.7% (a 'gzip --best' compression of the same data does only ~50%).
+* The list of all ~252K Unicode Code Points (32-bit numbers between 0 and 0x10ffff) is not structural trivial, but contains a lot of contiguous intervals. It can be compressed to a list of 700 integers, i.e. reduced by 99.7% (a 'gzip --best' compression of the same data achieves only ~50% compression).

-* There are 91 Fibonacci numbers greater than 1 fitting into 64 bit. This list is very hard to compress without knowledge of the special construction rule, because the numbers increase enormously fast. Here, IntZip achieves about 32% reduction by way of bit compression, wich is roughly the same reduction as obtained by 'gzip --best' applied to the same data.
+* There are 91 Fibonacci numbers greater than 1 fitting into 64 bit. This list is very hard to compress without knowledge of the special construction rule, because the numbers increase enormously fast. Here, _IntZip_ achieves about 30% reduction by way of bit compression, wich is roughly the same reduction as obtained by GNU gzip applied to the same data.

Home modified by Sebastian Böthin

Sebastian Böthin — Tue, 28 Apr 2015 21:29:43 -0000

--- v4
+++ v5
@@ -6,7 +6,7 @@

 * The complete list of all 65536 16-bit numbers has a trivial structure and can be compressed to just 6 bytes.

-* The list of all ~252K Unicode Code Points (32-bit numbers between 0 and 0x10ffff) doesn't have trivial structure, but contains a lot of contiguous intervals. It can be compressed to a list of ~14K 32-bit integers, i.e. reduced by 95%, while 'gzip --best' applied to the same data does only ~50%.
+* The list of all ~252K Unicode Code Points (32-bit numbers between 0 and 0x10ffff) doesn't have a trivial structure, but contains a lot of contiguous intervals. It can be compressed to a list of 700 32-bit integers, i.e. reduced by 99.7% (a 'gzip --best' compression of the same data does only ~50%).

-* There are 91 Fibonacci numbers greater than 1 and less than 2^64. This list is very hard to compress without knowledge of the special construction rule, because the numbers increase enormously fast. Here, IntZip achieves about 24% reduction by way of bit compression, wich is roughly the same reduction as obtained by 'gzip --best' when applied to the same data.
+* There are 91 Fibonacci numbers greater than 1 fitting into 64 bit. This list is very hard to compress without knowledge of the special construction rule, because the numbers increase enormously fast. Here, IntZip achieves about 32% reduction by way of bit compression, wich is roughly the same reduction as obtained by 'gzip --best' applied to the same data.

Home modified by Sebastian Böthin

Sebastian Böthin — Tue, 28 Apr 2015 21:29:42 -0000

--- v3
+++ v4
@@ -1,8 +1,8 @@
 _IntZip_ is a lossless compression algorithm for strictly increasing lists of unsigned integers.

-The algorithm compresses a suitable list of integers to a (whenever possible) shorter list of the same integer format. An arbitrary set of integers, i.e., a collection of unique numbers where the order does not matter, may be  transformed into suitable input data by way of translation and sorting.
+A suitable list of integers can be encoded to a (whenever possible) shorter list of the same integer format. An arbitrary set of integers, i.e., a collection of unique numbers where the order does not matter, may be  transformed into suitable input data by way of translation and sorting.

-The algorithm applies a combination of differential calculation, run-length encoding and bit compression (see  [Algorithm Description](Algorithm Description) for details). The quality of compression depends on diversity rather than the size of input. Therefore, _IntZip_ can produce on suitable integer lists in most cases much better compression results than general compression algorithms. For example:
+The algorithm applies a combination of differential calculation, run-length encoding and bit compression (see  [Algorithm Description](Algorithm Description) for details). The quality of compression depends on diversity rather than size of input. Therefore, _IntZip_ can produce on suitable integer lists in most cases much better compression results than general compression algorithms. For example:

 * The complete list of all 65536 16-bit numbers has a trivial structure and can be compressed to just 6 bytes.

Home modified by Sebastian Böthin

Sebastian Böthin — Tue, 28 Apr 2015 21:29:42 -0000

--- v2
+++ v3
@@ -2,7 +2,7 @@

 The algorithm compresses a suitable list of integers to a (whenever possible) shorter list of the same integer format. An arbitrary set of integers, i.e., a collection of unique numbers where the order does not matter, may be  transformed into suitable input data by way of translation and sorting.

-The algorithm applies a combination of differential calculation, run-length encoding and bit compression (see  [../Algorithm-Description](Algorithm Description) for details). The quality of compression depends on diversity rather than the size of input. Therefore, _IntZip_ can produce on suitable integer lists in most cases much better compression results than general compression algorithms. For example:
+The algorithm applies a combination of differential calculation, run-length encoding and bit compression (see  [Algorithm Description](Algorithm Description) for details). The quality of compression depends on diversity rather than the size of input. Therefore, _IntZip_ can produce on suitable integer lists in most cases much better compression results than general compression algorithms. For example:

 * The complete list of all 65536 16-bit numbers has a trivial structure and can be compressed to just 6 bytes.

Home modified by Sebastian Böthin

Sebastian Böthin — Tue, 28 Apr 2015 21:29:42 -0000

--- v1
+++ v2
@@ -1,8 +1,8 @@
-IntZip is a lossless compression algorithm for strictly increasing lists of unsigned integers.
+_IntZip_ is a lossless compression algorithm for strictly increasing lists of unsigned integers.

-The algorithm compresses a suitable list of unsigned integers to a (whenever possible) shorter list of the same integer format. That is, given a set of unsigned integers in 16, 32, or 64 bit size, the algorithm may be applied after the list has been sorted by value and provides a reduced list of the same bit size that can be efficently decompressed to the original list. Also efficient random access is possible, i.e., one can tell whether or not a particular integer is contained in the original list without the need of decompressing the whole list.
+The algorithm compresses a suitable list of integers to a (whenever possible) shorter list of the same integer format. An arbitrary set of integers, i.e., a collection of unique numbers where the order does not matter, may be  transformed into suitable input data by way of translation and sorting.

-The compression quality depends on the diversity of a list, rather than the size. Therefore, intzip can produce on increasing integer lists in most cases much better compression results than common compression algorithms. For example:
+The algorithm applies a combination of differential calculation, run-length encoding and bit compression (see  [../Algorithm-Description](Algorithm Description) for details). The quality of compression depends on diversity rather than the size of input. Therefore, _IntZip_ can produce on suitable integer lists in most cases much better compression results than general compression algorithms. For example:

 * The complete list of all 65536 16-bit numbers has a trivial structure and can be compressed to just 6 bytes.

Home modified by Sebastian Böthin

Sebastian Böthin — Tue, 28 Apr 2015 21:29:41 -0000

IntZip is a lossless compression algorithm for strictly increasing lists of unsigned integers.

The algorithm compresses a suitable list of unsigned integers to a (whenever possible) shorter list of the same integer format. That is, given a set of unsigned integers in 16, 32, or 64 bit size, the algorithm may be applied after the list has been sorted by value and provides a reduced list of the same bit size that can be efficently decompressed to the original list. Also efficient random access is possible, i.e., one can tell whether or not a particular integer is contained in the original list without the need of decompressing the whole list.

The compression quality depends on the diversity of a list, rather than the size. Therefore, intzip can produce on increasing integer lists in most cases much better compression results than common compression algorithms. For example:

The complete list of all 65536 16-bit numbers has a trivial structure and can be compressed to just 6 bytes.
The list of all ~252K Unicode Code Points (32-bit numbers between 0 and 0x10ffff) doesn't have trivial structure, but contains a lot of contiguous intervals. It can be compressed to a list of ~14K 32-bit integers, i.e. reduced by 95%, while 'gzip --best' applied to the same data does only ~50%.
There are 91 Fibonacci numbers greater than 1 and less than 2^64. This list is very hard to compress without knowledge of the special construction rule, because the numbers increase enormously fast. Here, IntZip achieves about 24% reduction by way of bit compression, wich is roughly the same reduction as obtained by 'gzip --best' when applied to the same data.