Method and apparatus for data filtering, and method and apparatus for constructing data filter

A method for constructing a data filter implemented by an apparatus, comprising:

• taking, by a processor of the apparatus, N sample vectors;
• for each sample vector of the N sample vectors:
• segmenting, by the processor, the sample vector to obtain k sample sub-vectors;
• respectively performing, by the processor, an inner product operation on the k sample sub-vectors and corresponding detection vectors among preset k detection vectors to obtain k second operation results;
• determining, by the processor, a second operation result, wherein a value of the second operation result is a maximum among the k second operation results;
• obtaining, by the processor, an identifier of a detection vector corresponding to the second operation result;
• and
• mapping, by the processor, the sample vector to a hash bucket corresponding to the identifier of the detection vector corresponding to the second operation result, wherein one hash bucket is preset for each detection vector, wherein a total of k hash buckets are preset, wherein the detection vector corresponding to the first operation result is in a one-to-one correspondence to an identifier, and wherein an identifier of a detection vector is also in a one-to-one correspondence to a hash bucket;
• respectively obtaining, by the processor, values of the k hash buckets to determine, according to a value of a corresponding hash bucket, whether to filter out the to-be-detected vector after the mapping of the N sample vectors is completed, wherein a value of each hash bucket is a total quantity of sample vectors that are mapped to the hash bucket;
• and
• determining, by the processor, that the construction of the data filter is completed, wherein the data filter comprises the k detection vectors, and wherein the k hash buckets that respectively correspond to values of corresponding hash buckets.

An apparatus for data filtering, comprising:

• a memory;
• and
• a processor, wherein the memory is configured to store an instruction needed by the processor to execute a program, and wherein the processor is configured to read the instruction stored in the memory to perform the following method:
• segmenting a to-be-detected vector to obtain k to-be-detected sub-vectors;
• respectively performing an inner product operation on the k to-be-detected sub-vectors and corresponding detection vectors among preset k detection vectors to obtain k first operation results;
• determining a first operation result, wherein a value of the first operation result is the maximum among the k first operation results;
• obtaining an identifier of a detection vector corresponding to the first operation result, wherein the detection vector corresponding to the first operation result is in a one-to-one correspondence to the identifier,
• mapping the to-be-detected vector to a preset data filter according to the identifier of the detection vector corresponding to the first operation result;
• and
• determining, using the preset data filter, whether to filter out the to-be-detected vector.

An apparatus for constructing a data filter, comprising:

• a processor;
• and
• a memory configured to store an instruction needed by the processor to execute a program, and wherein the processor is configured to read the instruction stored in the memory to perform:
• taking N sample vectors;
• for each sample vector of the N sample vectors:
• segmenting the sample vector to obtain k sample sub-vectors;
• respectively performing an inner product operation on the k sample sub-vectors and corresponding detection vectors among preset k detection vectors to obtain k second operation results;
• determining a second operation result, wherein a value of the second operation result is a maximum among the k first operation results;
• obtaining an identifier of a detection vector corresponding to the second operation result;
• mapping the sample vector to a hash bucket corresponding to the identifier of the detection vector corresponding to the second operation result, wherein one hash bucket is preset for each detection vector, wherein a total of k hash buckets are preset, wherein the detection vector corresponding to the second operation result is in a one-to-one correspondence to an identifier, and wherein an identifier of a detection vector is also in a one-to-one correspondence to a hash bucket;
• respectively obtain values of the k hash buckets to determine, according to a value of a corresponding hash bucket, whether to filter out the to-be-detected vector, after the mapping of the N sample vectors is completed wherein a value of each hash bucket is a total quantity of sample vectors that are mapped to the hash bucket;
• and
• determine that the construction of the data filter is completed, wherein the data filter comprises the k detection vectors, and wherein the k hash buckets that respectively correspond to values of corresponding hash buckets.