Method and apparatus for the multiplication of matrices and / or the multiplication of vectors

Abstract

The invention relates to a method for the multiplication of matrices and / or the multiplication of vectors, in particular for matrix-vector multiplication, having features of claim 1, and to an apparatus (30) for the multiplication of matrices and / or the multiplication of vectors, in particular for matrix-vector multiplication, having features of the additional independent claim.

Description

[0001] Method and apparatus for matrix multiplication and / or vector multiplication

[0002] Description

[0003] The invention relates to a method for matrix multiplication and / or vector multiplication, in particular matrix-vector multiplication, with features of claim 1, and to a device for matrix multiplication and / or vector multiplication, in particular matrix-vector multiplication, with features of the dependent claim. Optical pulses are used in this process. An optical pulse can be configured as a laser pulse. Photonic matrix-matrix or matrix-vector multiplication appears promising because it is more efficient and / or faster than available electronic solutions. Existing photonic architectures for addressing this problem have the significant disadvantage that their applicability is limited to matrix sizes of only a few rows or columns on the order of ten by ten.The reason for this limitation is the fact that the multiplication of the matrix elements usually takes place in the space domain. This means that a single physical element (such as a ring resonator and others) is used for each multiplication.

[0004] Furthermore, in some cases wavelength-division multiplexing must also be used to avoid unwanted mutual coupling between these functional elements. This limits the size of the matrices that can be processed in one step to the size of the physical implementation (i.e., the number of individual elements). Moreover, it prevents the dimensions of "space" and "wavelength" from being freely used to multiply the computing capacity of the implementation.

[0005] It is therefore an object of the present invention to provide a method and an apparatus for matrix multiplication and / or vector multiplication, whereby the above disadvantages are eliminated. The above object is achieved by a method for matrix multiplication and / or vector multiplication, in particular for matrix-vector multiplication, with the features of claim 1.

[0006] The procedure includes the following steps:

[0007] Providing and / or generating at least one first sequence of optical pulses, wherein each optical pulse of the first sequence represents a different element of a first matrix or a first vector.

[0008] Providing and / or generating at least one second sequence of optical pulses, wherein the individual optical pulses of the second sequence each represent a different element of a second matrix or a second vector.

[0009] Generating at least one third sequence of optical pulses by introducing the first and second sequences of optical pulses into a nonlinear medium, wherein in the medium each optical pulse from the first sequence interacts with each optical pulse from the second sequence, generating each optical pulse of the third sequence whose optical field strength amplitude corresponds to the product of the optical field strength amplitudes of the two optical pulses from the first and second sequences. Each optical pulse of the third sequence can be used to determine a different element of a third matrix or a third vector. Detecting the individual optical pulses of the third sequence.

[0010] Following detection, a number of optical pulses of the third sequence can be summed, corresponding to the size of the row or column vectors of the original (first and second) matrices or vectors. This sum of the optical pulses of the third sequence yields the corresponding element of the result matrix or vector (or the third matrix or vector), according to the rules of mathematical matrix multiplication.

[0011] The nonlinear medium is a medium whose optical response function contains nonlinear components of at least the second, third, fourth and / or higher order.

[0012] Thus, matrix multiplication can be performed entirely in a time domain by generating temporal sequences of optical pulses. This can be achieved by transforming the matrices into a sequential sequence of their row and column vectors. This transformation can be performed in an element with second-order nonlinearity for multiplication. The resulting optical pulses with higher-order frequencies are detected by a detector at the end of the chain.

[0013] The multiplication can be performed with any matrix shape and size, since the matrices are converted into temporal sequences of optical pulses (or optical pulse streams) whose duration need not be limited. According to a further development of the method, the procedure can include the following step:

[0014] Filtering out unwanted frequencies from the third sequence of optical pulses, so that only the products of the optical pulses of the first and second sequences required for determining the matrix elements are passed on to the detector.

[0015] This allows the detection of unwanted optical signals to be prevented or at least reduced using simple means.

[0016] According to a further development of the procedure, the procedure can include the following step:

[0017] Modulating an intensity and / or an amplitude of the optical field strength of the first sequence of optical pulses such that the value of the intensity and / or the amplitude of the optical field strength of the respective optical pulses represents the value of a different element of the first matrix or the first vector.

[0018] Alternatively or additionally, modulating the intensity and / or amplitude of the optical field strength of the second sequence of optical pulses such that the value of the intensity and / or amplitude of the optical field strength of the respective optical pulses corresponds to the value of a different element of the second matrix or the second vector. Alternatively or additionally, detecting the intensity and / or amplitude of the optical field strength of the third sequence of optical pulses.

[0019] This allows the respective element of a matrix or vector to be converted or encoded into an optical pulse using simple means.

[0020] According to a further development of the procedure, the procedure can include the following step:

[0021] Use and / or utilization (or generation) of a waiting period after a predetermined number of optical pulses of the first and / or the second sequence to generate a block of optical pulses, wherein the block of optical pulses represents a single vector, in particular a column vector and / or a row vector.

[0022] This allows for the separation into individual vectors using simple means.

[0023] According to a further development of the procedure, the procedure can include the following step:

[0024] Storing the elements of the third matrix or vector represented by the optical pulses of the third sequence into a memory, in particular a computer-readable data storage device, and in particular generating the third matrix or third vector from the stored elements. This allows the optical pulses of the third sequence or the elements represented by the optical pulses of the third sequence to be subsequently processed, e.g., summed. The processing of the stored optical pulses of the third sequence or the elements represented by the respective optical pulses can, for example, be implemented electronically. In particular, the memory or the elements stored in the memory can be accessed (at any time).

[0025] According to a further development of the procedure, the procedure can include the following step:

[0026] Quadrature amplitude modulation of the first sequence of optical pulses such that the value of an amplitude of the optical field strength and a phase of the individual optical pulses of the first sequence represents the value of a different element of the first matrix or the first vector.

[0027] Alternatively or additionally, quadrature amplitude modulation of the second sequence of optical pulses such that the value of an amplitude of the field strength and a phase of the individual optical pulses represents the value of a different element of the second matrix or the second vector.

[0028] By using the amplitude and phase, the assignment of the individual elements of the matrices or vectors to the respective optical pulses can be further optimized. According to a further development of the method, the procedure can comprise the following steps:

[0029] Providing and / or generating at least two first sequences of optical pulses, wherein the at least two first sequences of optical pulses each have a different frequency.

[0030] Providing and / or generating at least two second sequences of optical pulses, wherein the at least two second sequences of optical pulses each have a different frequency.

[0031] Generating at least two third sequences of optical pulses, each by introducing one of the at least two first sequences of optical pulses and one of the at least two second sequences of optical pulses into the nonlinear medium, wherein the at least two third sequences of optical pulses each have a different frequency.

[0032] This allows, for example, the generation of two third sequences of optical pulses by creating two first sequences and two second sequences of optical pulses. One of the first two sequences of optical pulses can interact with the first of the two second sequences of optical pulses in the nonlinear medium, generating the first of the two third sequences of optical pulses. Similarly, a second of the first two sequences of optical pulses can interact with the second of the two second sequences of optical pulses in the nonlinear medium, generating the second of the two third sequences of optical pulses. This enables multiple multiplications to be performed on the same nonlinear medium, particularly simultaneously. In particular, it allows for scaling up the multiplication while maintaining the same (or similar) space requirements.In particular, the capacity can be multiplied by using several wavelengths with the same central components or the same nonlinear medium.

[0033] According to a further development of the method, the first sequence of optical pulses and the second sequence of optical pulses can have the same repetition rates and interact simultaneously in the nonlinear medium.

[0034] Alternatively or additionally, the optical pulses of the first sequence and the optical pulses of the second sequence can have the same pulse durations.

[0035] Alternatively or additionally, the optical pulses of the first sequence and the optical pulses of the second sequence can have different frequencies and / or wavelengths.

[0036] This allows for optimal interaction of the respective optical pulses in the nonlinear medium using simple means.

[0037] According to a further development of the procedure, the procedure can include the following step:

[0038] Adding the individual optical pulses of the third sequence. This can be implemented electronically, for example, after the detection of the optical pulses. This allows the third matrix to be reconstructed in the usual matrix notation.

[0039] The above problem is solved by a device for matrix multiplication and / or vector multiplication, in particular for matrix-vector multiplication, with the features of the dependent claim. The device is configured to carry out the method according to the above descriptions.

[0040] Regarding the advantages achievable with the device, reference is made to the relevant explanations of the method. The measures described in connection with the method and / or those explained below can be used for further development of the device.

[0041] According to a further development of the device, the device can comprise at least one laser. The first laser can be configured to generate the first sequence of optical pulses. The device can comprise at least one second laser. The second laser can be configured independently or separately from the first laser. The second laser can be configured to generate the second sequence of optical pulses.

[0042] This allows the first and second sequences of optical pulses to be generated using simple means. According to a further development of the device, the device can include at least one filter for filtering out unwanted frequencies from the individual optical pulses of the third sequence. The filter can be configured as a high-pass filter. The unwanted frequencies can be, in particular, the frequency of the first sequence of optical pulses, the frequency of the second sequence of optical pulses, and / or the subtraction of the frequencies of the first and second sequences of optical pulses.

[0043] This allows unwanted frequencies and thus interference signals or interference pulses to be filtered out or at least reduced using simple means.

[0044] According to a further development of the device, the device can comprise at least one detector for detecting the individual optical pulses of the third sequence. The detector can be configured as a photodetector. The detector can be configured as an analog integrator. The effective bandwidth of the detector can, in particular, be lower than the bandwidth of the individual optical pulses of the third sequence.

[0045] According to a further development, the device can include a photonic integrated circuit. The nonlinear medium, the first laser, the second laser and / or the detector can be arranged on the photonic integrated circuit.

[0046] This allows for a highly compact device design. According to a further development of the device, at least two nonlinear media, at least two first lasers, at least two second lasers, and / or at least two detectors can be arranged on the photonic integrated circuit. Thus, at least two matrix and / or vector multiplications can be performed, particularly simultaneously.

[0047] The photonic integrated circuit can be a lithium niobate (LiNbOS) chip. The photonic integrated circuit can form a component of a LiNbOS chip.

[0048] This makes spatial scaling conceivable using simple means. For example, multiple implementations (matrix and / or vector multiplications) can be implemented side-by-side on the same photonic integrated circuit or on the same LiNbOS chip.

[0049] According to a further development of the device, the device can comprise at least two first lasers. The at least two first lasers can each be configured to generate a first sequence of optical pulses. The at least two first sequences of optical pulses can each have a different frequency.

[0050] Alternatively or additionally, the device can comprise at least two second lasers. Each of the at least two second lasers can be configured to generate a second sequence of optical pulses. Each of the at least two second sequences of optical pulses can have a different frequency.

[0051] Alternatively or additionally, the device can include at least two filters. The filters can be configured to filter out unwanted frequencies from the optical pulses of the at least two third sequences.

[0052] Alternatively or additionally, the device can comprise at least two detectors. The detectors can be configured to detect the individual optical pulses of the at least two third sequences.

[0053] Alternatively or additionally, the device can have at least one filter element for filtering out different frequencies and forwarding them to the corresponding detectors.

[0054] Alternatively or additionally, the device can have at least one element for combining the different frequencies on the input side, in particular a wavelength multiplexer and / or a wavelength demultiplexer.

[0055] All devices mentioned previously and / or subsequently can each also constitute a partial step of the method according to the invention.

[0056] This allows multiple multiplications, especially simultaneous ones, to be performed on the same nonlinear medium. In particular, it enables scaling up the multiplication while maintaining the same (or similar) space requirements. For example, the capacitance can be multiplied by using multiple wavelengths with the same central components or the same nonlinear medium.

[0057] Furthermore, the invention may include the use of the method and / or the device according to the invention within the framework of a neural network, in particular within the framework of training the neural network.

[0058] Further features, details and advantages of the invention will become apparent from the wording of the claims and from the following description of exemplary embodiments with reference to the drawings. The drawings show:

[0059] Fig. 1 shows a diagram of a procedure for matrix multiplication and / or vector multiplication and

[0060] Fig. 2 shows a schematic representation of a device for carrying out the method according to Figure 1.

[0061] Figure 1 shows a diagram of a matrix multiplication and / or vector multiplication procedure. The diagram is intended solely to graphically illustrate the relationships between the individual steps of the procedure. In particular, the diagram does not depict a temporal sequence or dependency of the steps shown.

[0062] The procedure includes the following steps:

[0063] 10: Providing and / or generating at least one first sequence of optical pulses 11 (see Figure 2). The individual optical pulses of the first sequence each represent a different element of a first matrix or a first vector.

[0064] 12: Providing and / or generating at least one second sequence of optical pulses 13 (see Figure 2). The individual optical pulses of the second sequence each represent a different element of a second matrix or a second vector.

[0065] 14: Generating at least one third sequence of optical pulses 15 (see Figure 2) by introducing the first and second sequences of optical pulses 11, 13 into a nonlinear medium 17. The nonlinear medium can, in particular, be a nonlinear medium of second, third, fourth, or higher order. In the medium 17, each optical pulse from the first sequence of optical pulses 11 interacts with each optical pulse from the second sequence of optical pulses 13 and generates each optical pulse of the third sequence of optical pulses 15. Each of the individual optical pulses of the third sequence can be used to determine another element of a third matrix or a third vector.

[0066] 16: Detecting the individual optical pulses of the third sequence.

[0067] Following detection, a number of optical pulses of the third sequence can be summed, corresponding to the size of the row or column vectors of the original matrices or vectors. This sum of the optical pulses of the third sequence yields the corresponding element of the result matrix or vector (or the third matrix or vector), according to the rules of mathematical matrix multiplication.

[0068] The procedure may include the following step:

[0069] 18 : From filtering out unwanted frequencies from the third sequence of optical pulses 15 , such that the individual optical pulses of the third sequence correspond to a multiplication of the respective elements represented by the respective interacting optical pulses of the first sequence and the second sequence .

[0070] The procedure may include the following step:

[0071] 20 : Modulating an intensity and / or an amplitude of the optical field strength of the first sequence of optical pulses 11 such that the value of the intensity and / or the amplitude of the optical field strength of the respective optical pulses represents the value of a respective different element of the first matrix or the first vector .

[0072] Alternatively or additionally, modulating the intensity and / or amplitude of the optical field strength of the second sequence of optical pulses 13 such that the value of the intensity and / or amplitude of the optical field strength of the respective optical pulses corresponds to the value of a different element of the second matrix or the second vector. Alternatively or additionally, detecting the intensity and / or amplitude of the optical field strength of the third sequence of optical pulses 15.

[0073] The procedure may include the following step:

[0074] 22 : Use and / or utilization (or generation) of a predetermined waiting time after a predetermined number of optical pulses of the first and second sequences to generate a block of optical pulses each, wherein the block of optical pulses represents a single vector, in particular a column vector and / or row vector.

[0075] The procedure may include the following step:

[0076] 24: Storing the elements of the third matrix or the third vector represented by the optical pulses of the third sequence in a memory, and in particular generating the third matrix or the third vector from the stored elements. The storage of the optical pulses of the third sequence can include immediate summation, in particular immediate block summation. Alternatively or additionally, the storage of the information regarding the optical pulses of the third sequence can first involve intermediate storage followed by summation, in particular subsequent block summation, of the stored information regarding the optical pulses of the third sequence.

[0077] The procedure may include the step: 26 : Quadrature amplitude modulation of the first sequence of optical pulses 11 such that the value of an amplitude of the optical field strength and a phase of the individual optical pulses of the first sequence represents the value of a different element of the first matrix or the first vector.

[0078] Alternatively or additionally, quadrature amplitude modulation of the second sequence of optical pulses 13 such that the value of an amplitude of the optical field strength and a phase of the individual optical pulses represents the value of a different element of the second matrix or the second vector.

[0079] Alternatively or additionally, detection of a quadrature amplitude modulation of the optical field strength of the third sequence of optical pulses 15 .

[0080] The process may include the following steps:

[0081] 28: Providing and / or generating at least two first sequences of optical pulses 11, wherein the at least two first sequences of optical pulses 11 each have a different frequency and

[0082] Providing and / or generating at least two second sequences of optical pulses 13, wherein the at least two second sequences of optical pulses 13 each have a different frequency and

[0083] Generating at least two third sequences of optical pulses 15, each by introducing one of the at least two first sequences of optical pulses 11 and one of the at least two second sequences of optical pulses 13 into the nonlinear medium 17, wherein the at least two third sequences of optical pulses 15 each have a different frequency.

[0084] The first sequence of optical pulses 11 and the second sequence of optical pulses 13 can have the same repetition rates and interact simultaneously in the nonlinear medium 17.

[0085] Alternatively or additionally, the optical pulses of the first episode 11 and the optical pulses of the second episode 13 can have the same pulse durations.

[0086] Alternatively or additionally, the optical pulses of the first episode 11 and the optical pulses of the second episode 13 can have different frequencies and / or wavelengths.

[0087] The procedure may include the following step:

[0088] 29: Summing the individual optical pulses of the third sequence 15. For example, the values ​​or elements stored in step 24 can be summed in an electronic process. Alternatively or additionally, the addition can be performed in a detector, in particular an integrating detector.

[0089] Figure 2 shows a schematic representation of a device 30 for carrying out the method according to Figure 1. The device 30 shown is configured for matrix multiplication and / or vector multiplication, in particular for matrix-vector multiplication.

[0090] The device 30 can comprise at least one first laser 32. The first laser 32 can be configured to generate the first sequence of optical pulses 11.

[0091] The device 30 can comprise at least one second laser 34. The second laser 34 can be configured to generate the second sequence of optical pulses 13. The second laser 34 is, in particular, independent of the first laser 32. The two lasers 32, 34 can be configured separately and / or spaced apart from each other.

[0092] The device 30 can comprise at least one filter 36. The filter 36 can be configured to filter out unwanted frequencies from the individual optical pulses of the third sequence. The filter 36 can, in particular, be configured as a high-pass filter.

[0093] The device 30 can comprise at least one detector 38. The detector 38 can be configured to detect the individual optical pulses of the third sequence. The detector 38 can be configured as a photodetector.

[0094] The device 30 can comprise a photonic integrated circuit 40. The nonlinear medium 17, the first laser 32, the second laser 34, and / or the detector 38 can be arranged on the photonic integrated circuit 40. In Figure 2, only the nonlinear medium 17 is shown on the photonic integrated circuit 40. At least two nonlinear media 17, at least two first lasers 32, at least two second lasers 34, and / or at least two detectors 38 can be arranged on the photonic integrated circuit 40, enabling at least two matrix and / or vector multiplications to be performed. This can be implemented, in particular, simultaneously.

[0095] The device 30 can comprise at least two first lasers 32. The at least two first lasers 32 can each be configured to generate a first sequence of optical pulses 11. The at least two first sequences of optical pulses 11 can each have a different frequency.

[0096] Alternatively or additionally, the device 30 can comprise at least two second lasers 34. The at least two second lasers 34 can each be configured to generate a second sequence of optical pulses 13. The at least two second sequences of optical pulses 13 can each have a different frequency.

[0097] Alternatively or additionally, the device 30 can have at least one element for combining the different frequencies on the input side, in particular a wavelength multiplexer and / or a wavelength demultiplexer.

[0098] Alternatively or additionally, the device can comprise at least two filters 36. The at least two filters 36 can each be configured to filter out unwanted frequencies from the optical pulses of the at least two third sequences.

[0099] Alternatively or additionally, the device can comprise at least two detectors 38. The at least two detectors 38 can be configured to detect the individual optical pulses of the at least two third sequences.

Claims

Patent claims 1. Methods for matrix multiplication and / or vector multiplication, in particular for matrix-vector multiplication, comprising the steps: (10) Providing and / or generating at least one first sequence of optical pulses (11) , wherein each optical pulse of the first sequence represents a different element of a first matrix or vector; (12) Providing and / or generating at least one second sequence of optical pulses (13) , wherein each optical pulse of the second sequence represents a different element of a second matrix or vector; (14) Generating at least one third sequence of optical pulses (15) by introducing the first and second sequences of optical pulses (11, 13) into a nonlinear medium (17) , wherein in the medium (17) each optical pulse from the first sequence of optical pulses (11) interacts with an optical pulse from the second sequence of optical pulses (13) and each generates an optical pulse of the third sequence of optical pulses (15), wherein each of the individual optical pulses of the third sequence can be used to determine another element of a third matrix or a third vector; (16) Detecting the individual optical pulses of the third sequence.

2. The method according to claim 1, characterized by the step: (18) Filtering out unwanted frequencies from the third sequence of optical pulses ( 15 ) , such that the individual optical pulses of the third sequence correspond to a multiplication of the respective elements represented by the respective interacting optical pulses of the first sequence and the second sequence, which are used to determine the elements of the third matrix or the third vector .

3. Method according to claim 1 or claim 2, characterized by the step: (20) Modulating an intensity and / or an amplitude of the optical field strength of the first sequence of optical pulses (11) such that the value of the intensity and / or the amplitude of the optical field strength of each optical pulse represents the value of a different element of the first matrix or vector and / or - Modulating an intensity and / or an amplitude of the optical field strength of the second sequence of optical pulses ( 13 ) such that the value of the intensity and / or the The amplitude of the optical field strength of the respective optical pulses represents the value of a different element of the second matrix or the second vector and / or - Detection of an intensity and / or an amplitude of the optical field strength of the third sequence of optical pulses ( 15 ).

4. Method according to one of the preceding claims, characterized by the step: (22) Use and / or utilization of a waiting period after a predetermined number of optical pulses of the first and / or the second sequence to generate a block of optical pulses, wherein the block of optical Pulse represents a single vector, in particular a column vector and / or a row vector.

5. Method according to one of the preceding claims, characterized by the step: ( 24 ) Storing the elements of the third matrix or vector generated by the optical pulses of the third sequence into a memory and / or generating the third matrix or vector from the stored elements .

6. Method according to one of the preceding claims, characterized by the step: (26) Quadrature amplitude modulation of the first sequence of optical pulses (11) such that the value of an amplitude of the optical field strength and a phase of the individual optical pulses of the first sequence represents the value of a different element of the first matrix or the first vector and / or - Quadrature amplitude modulation of the second sequence of optical pulses ( 13 ) such that the value of an amplitude of the optical field strength and a phase of the individual optical pulses represents the value of a respective different element of the second matrix or the second vector , and / or - Detection of the quadrature amplitude modulation of the optical field strength of the third sequence of optical pulses ( 15 ).

7. Method according to one of the preceding claims, characterized by the steps: (28) Providing and / or generating at least two first sequences of optical pulses (11) wherein the at least two first sequences of optical pulses (11) each have a different frequency and - Providing and / or generating at least two second sequences of optical pulses (13) , wherein the at least two second sequences of optical pulses (13) each have a different frequency and - Generating at least two third sequences of optical pulses (15) by introducing one of the at least two first sequences of optical pulses (11) and one of the at least two second sequences of optical pulses (13) into the nonlinear medium (17), wherein the at least two third sequences of optical pulses (15) each have a different frequency.

8. Method according to one of the preceding claims, characterized in that - the first sequence of optical pulses (11) and the second sequence of optical pulses (13) have the same repetition rates and interact simultaneously in the nonlinear medium (17), and / or - the optical pulses of the first sequence (11) and the optical pulses of the second sequence (13) have the same pulse durations, and / or - the optical pulses of the first sequence (11) and the optical pulses of the second sequence (13) have different frequencies and / or wavelengths.

9. Method according to one of the preceding claims, characterized by the step: (29) Adding up the individual optical pulses of the third sequence (15) .

10. Device (30) for matrix multiplication and / or vector multiplication, in particular for matrix-vector multiplication, wherein the device (30) is configured to carry out the method according to one of the preceding claims.

11. Device (30) according to claim 10, characterized in that the device (30) comprises: - at least one first laser (32) , wherein the first laser (32) is configured to generate the first sequence of optical pulses (11), - at least one second laser (34), in particular independent of the first laser (32), wherein the second laser (34) is configured to generate the second sequence of optical pulses (13).

12. Device (30) according to claim 10 or 11, characterized in that the device (30) comprises at least one filter (36), in particular a high-pass filter, for filtering out unwanted frequencies from the individual optical pulses of the third sequence.

13. Device (30) according to one of claims 10 to 12, characterized in that the device (30) comprises at least one detector (38) for detecting the individual optical pulses of the third sequence.

14. Device (30) according to one of claims 10 to 13, characterized in that the device (30) comprises a photonic integrated circuit (40), wherein the nonlinear medium (17), the first laser (32), the second laser (34) and / or the detector (38) are arranged on the photonic integrated circuit (40).

15. Device (30) according to the preceding claim, characterized in that at least two nonlinear media (17), at least two first lasers (32), at least two second lasers (34) and / or at least two detectors (38) are arranged on the photonic integrated circuit (40) so that at least two matrix and / or vector multiplications, in particular simultaneously, can be performed.

16. Device according to one of claims 10 to 15, characterized in that the device (30) comprises: - at least two first lasers (32) , wherein the at least two first lasers (32) are each configured to generate a first sequence of optical pulses (11), wherein the at least two first sequences of optical pulses (11) each have a different frequency, - at least two second lasers (34) , wherein the at least two second lasers (34) are each configured to generate a second sequence of optical pulses (13), wherein the at least two second sequences of optical pulses (13) each have a different frequency, - at least two filters (36) , wherein the filters (36) are configured to filter out unwanted frequencies from the optical pulses of the at least two third sequences and / or - at least two detectors (38) , wherein the detectors (38) are configured to detect the individual optical pulses of the at least two third sequences.

Images