Method and system for automatic selection of model parameters based on hidden markov

Abstract

The application discloses a model parameter automatic selection method and system based on a hidden Markov model. The method comprises the following steps: in response to a selection instruction input by a user in a visual operation interface, a selected region in the model is determined, and a selected region set is constructed; a hidden state set is constructed, wherein the hidden state set contains adjustable parameters of the model; a state transition probability matrix and an observation probability matrix are obtained; a hidden state sequence is calculated based on the selected region, the obtained state transition probability matrix and the observation probability matrix by using a Viterbi algorithm to obtain a predicted parameter combination. The technical scheme of the application fixes expert experience in the state transition matrix and the observation probability matrix based on the hidden Markov algorithm. When a user selects certain regions, a parameter combination is automatically predicted and prompted, the user is guided, the experience threshold for use is reduced, and the efficiency of parameter selection by technical personnel is improved.

Description

Technical Field

[0001] This invention belongs to the field of computer-aided design technology for integrated circuits, and particularly relates to a method and system for automatic selection of model parameters based on Hidden Markov Models. Background Technology

[0002] The continuous development of semiconductor and integrated circuit technologies has made integrated circuit computer-aided design (CAD) or electronic design automation (EDA) platforms increasingly important. A fundamental function of an EDA platform is parameter extraction from device models, that is, extracting model parameters of semiconductor devices manufactured using specific integrated circuit processes based on standard device models. After the model parameters are extracted, combined with the corresponding standard device models, various operating characteristics of the semiconductor device can be mathematically described, thus enabling device simulation in subsequent circuit design.

[0003] The BSIM model is a metal-oxide-semiconductor field-effect transistor (MOSFET) model developed by the University of California, Berkeley, suitable for digital and analog circuit design and simulation. In practical parameter extraction, various BSIM models corresponding to the actual MOSFET device (such as BSIM4, BSIM-Bulk, BSIM-CMG, etc.) can be selected to process the MOSFET device's test data (e.g., IV curves and CV curves of MOSFETs of different sizes), and the model parameters of the MOSFET device can then be extracted.

[0004] Current techniques for extracting parameters from various device models involve first selecting a region to be fitted, then manually inputting parameters based on experience, and finally modifying these parameters through optimization algorithms to fit the selected region. However, this parameter selection method relies too heavily on experience and is unsuitable for personnel lacking experience. Summary of the Invention

[0005] To address the aforementioned problems, the present invention aims to provide an automatic parameter selection method and system based on Hidden Markov Models. This method and system, based on the Hidden Markov algorithm, embeds expert experience into the state transition matrix and observation probability matrix. When a user selects certain regions, the system automatically predicts and prompts for parameter combinations, guiding the user and reducing the experience threshold for users, thereby improving the efficiency of parameter selection by technical personnel.

[0006] The first technical solution provided in this application is: an automatic parameter selection method for a Hidden Markov Model, comprising the following steps: responding to a selection command input by a user in a visual operation interface to determine a selected region in the model, and constructing a set of selected regions; constructing a set of hidden states, wherein the set of hidden states contains adjustable parameters of the model; obtaining a state transition probability matrix and an observation probability matrix; and using the Viterbi algorithm to calculate a sequence of hidden states based on the selected regions and the obtained state transition probability matrix and observation probability matrix to obtain a predicted parameter combination.

[0007] Preferably, in response to a selection command input by the user in the visualization interface, determining a selected region in the model and constructing a set of selected regions further includes: obtaining attribute information of the representation graphic selected by the cursor in the current visualization interface; obtaining parameter information of all points in the representation graphic that are in a marked state; determining the corresponding selected region in the model based on the attribute information of the representation graphic and the parameter information of all points in the representation graphic that are in a marked state; and constructing a set of selected regions V = {v1, v2, ..., v...} m}, m represents the number of possible observations.

[0008] Preferably, constructing a set of hidden states, wherein the set of hidden states includes adjustable parameters of the model, further comprises: constructing a set of hidden states Q = {q1, q2, ..., q...} n}, where n represents the number of possible states; the model includes, but is not limited to, the following models: BSIM3, BSIM4, BSIM6, BSIM-CMG, BSIM-IMG, BSIMSOI, UTSOI, HiSIM2, HiSIM_HV, PSP, GP-BJT, RPITFT, and each state q in the hidden state set. n These correspond to the adjustable parameters in the model.

[0009] Preferably, constructing a set of hidden states, wherein the set of hidden states includes adjustable parameters of the model, further includes: each state q in the set of hidden states n The adjustable parameter corresponds to one or a combination of the adjustable parameters, wherein the adjustable parameter is the adjustable parameter in the model.

[0010] Preferably, the method further includes, after calculating the hidden state sequence using the Viterbi algorithm, performing deduplication processing on the hidden state sequence to obtain the predicted parameter combination.

[0011] Preferably, the state transition probability matrix A is: A = [a ij ] N×N , where a ij =P(it+1 =q j |i t =q i ), i=1, 2,…,N; j=1, 2,…,N; a ij Represents the state at time t in g i State, transition to q at time t+1 j The probability of a state.

[0012] Preferably, the observation probability matrix B is: B = [b j (k)] N×M , where b j (k)=P(o t =v k |i t =q j ), k=1, 2,…,M; j=1, 2,…,N; b j (k) represents the state at time t, q i State generation observation v k The probability of generation is also called the probability of emission.

[0013] Preferably, the method of calculating the hidden state sequence using the Viterbi algorithm based on the selected region and the obtained state transition probability matrix and observation probability matrix to obtain the predicted parameter combination further includes: determining the observation sequence O = (o1, o2, ... o3) based on the selected region. T ), where T is the number of selected regions; combining the state transition probability matrix A, the observation probability matrix B, and the number of selected regions T, the Viterbi algorithm is used to calculate k hidden state sequences that may generate observation event sequences, so as to obtain the predicted k sets of parameter combinations.

[0014] Based on the same concept, this invention also provides an automatic parameter selection system for Hidden Markov Models, comprising: a selected region acquisition module, used to determine a selected region in the model in response to a selection command input by a user in a visual operation interface, and construct a set of selected regions; a parameter combination construction module, used to construct a set of hidden states, wherein the set of hidden states contains adjustable parameters of the model; a matrix acquisition module, used to acquire a state transition probability matrix and an observation probability matrix; and a prediction module, used to calculate a sequence of hidden states using the Viterbi algorithm based on the selected regions and the acquired state transition probability matrix and observation probability matrix to obtain the predicted parameter combination.

[0015] Based on the same concept, the present invention also provides an electronic device, characterized in that it includes: a memory for storing a processing program; and a processor, which, when executing the processing program, implements the automatic selection method for model parameters based on Hidden Markov Models as described above.

[0016] Based on the same concept, the present invention also provides a readable storage medium, characterized in that the readable storage medium stores a processing program, which, when executed by a processor, implements the automatic selection method for model parameters based on Hidden Markov as described above.

[0017] Because the present invention adopts the above technical solution, it has the following advantages and positive effects compared with the prior art:

[0018] 1. The technical solution of this embodiment is based on the Hidden Markov Model (HMM) to predict the combination of selected parameters according to the selected region. First, the expert experience is solidified in the state transition matrix and the observation probability matrix. When the user selects a region, the system will automatically predict and prompt the parameter combination based on the constructed HMM and the state transition matrix and observation probability matrix solidified by the expert experience, guide the user, reduce the experience threshold for the user, and improve the efficiency of parameter selection by technical personnel.

[0019] 2. The state transition probability matrix A and observation probability matrix B mentioned in the technical solution of this embodiment are means of solidifying expert experience. By solidifying expert experience and combining it with the hidden Markov algorithm, scientific prediction of parameter combinations can be achieved, automatically predicting and prompting parameter combinations, guiding users, reducing the experience threshold for users, and improving the efficiency of parameter selection by technical personnel. Attached Figure Description

[0020] The specific embodiments of the present invention will be further described in detail below with reference to the accompanying drawings, wherein:

[0021] Figure 1 This is a flowchart of the method for automatically selecting model parameters based on Hidden Markov Models in this invention.

[0022] Figure 2 This is a schematic diagram of the ID-Vg characterization of the present invention;

[0023] Figure 3 This is a schematic diagram of the Vtlin characterization of the present invention. Detailed Implementation

[0024] The present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments. The advantages and features of the present invention will become clearer from the following description and claims. It should be noted that the drawings are all in a very simplified form and use non-precise ratios, and are only used to facilitate and clarify the illustration of the embodiments of the present invention.

[0025] It should be noted that all directional indications (such as up, down, left, right, front, back, etc.) in the embodiments of the present invention are only used to explain the relative positional relationship and movement of each component in a certain specific posture (as shown in the figure). If the specific posture changes, the directional indication will also change accordingly.

[0026] First Embodiment

[0027] like Figure 1 As shown, this embodiment provides the following technical solution: an automatic parameter selection method for a Hidden Markov Model, comprising the following steps:

[0028] S100: In response to a selection command input by the user in the visual operation interface, determine the selected region in the model and construct a set of selected regions;

[0029] S200: Construct a set of hidden states, wherein the set of hidden states contains adjustable parameters of the model;

[0030] S300: Obtain the state transition probability matrix and observation probability matrix;

[0031] S400: Based on the selected region and the obtained state transition probability matrix and observation probability matrix, the Viterbi algorithm is used to calculate the hidden state sequence to obtain the predicted parameter combination.

[0032] The technical solution of this embodiment is based on the Hidden Markov Model (HMM) to predict the combination of selected parameters according to the selected region. First, the expert experience is solidified in the state transition matrix and the observation probability matrix. When the user selects a region, the system will automatically predict and prompt the parameter combination based on the constructed HMM and the state transition matrix and observation probability matrix solidified by the expert experience, guide the user, reduce the experience threshold for the user, and improve the efficiency of parameter selection by technical personnel.

[0033] Preferably, in response to a selection command input by the user in the visualization interface, determining a selected region in the model and constructing a set of selected regions further includes: obtaining attribute information of the representation graphic selected by the cursor in the current visualization interface; obtaining parameter information of all points in the representation graphic that are in a marked state; determining the corresponding selected region in the model based on the attribute information of the representation graphic and the parameter information of all points in the representation graphic that are in a marked state; and constructing a set of selected regions V = {v1, v2, ..., v...} m}, m represents the number of possible observations.

[0034] In the technical solution of this embodiment, a method for obtaining a selected region is provided, see [link to relevant documentation]. Figure 2 and Figure 3 , Figure 2All solid dots in ID-Vg represent the selected area. Figure 3 In a Vtlin graph, a point in this kop can also be considered a selected region. The set of all selected regions, V, is the set of visible states, where V = {v1, v2, ..., vm}, and m represents the number of possible observations. The selected regions are chosen by the user.

[0035] Preferably, constructing a set of hidden states, wherein the set of hidden states includes adjustable parameters of the model, further comprises: constructing a set of hidden states Q = {q1, q2, ..., q...} n}, where n represents the number of possible states; the model includes, but is not limited to, the following models: BSIM3, BSIM4, BSIM6, BSIM-CMG, BSIM-IMG, BSIMSOI, UTSOI, HiSIM2, HiSIM_HV, PSP, GP-BJT, RPITFT, and each state q in the hidden state set. n These correspond to the adjustable parameters in the model.

[0036] For example, taking the BSIM6 model as an example, the parameters in the model include common adjustable parameters such as a0, a1, vth0, voff, nfactor, k2, and cit. Of course, it can also include other parameters not listed here; this is only an example. The implicit states mentioned here refer to the parameters or combinations of parameters, for example, q1 = {a0}, q2 = {a1}, q3 = {vth0}, q4 = {voff}, etc.

[0037] Preferably, the devices used in the model of this embodiment include, but are not limited to, MOSFET transistors, silicon-on-insulator (SOI) transistors, FinFETs, bipolar junction transistors (BJTs), heterojunction transistors (HBTs), thin-film transistors (TFTs), metal-semiconductor contact field-effect transistors (MESFETs), diodes, resistors, or inductors. The determined device models can be, but are not limited to, BSIM3, BSIM4, BSIM6, BSIM-CMG, BSIM-IMG, BSIMSOI, UTSOI, HiSIM2, HiSIM_HV, PSP, GP-BJT, or RPITFT. For example, for a MOSFET transistor, the corresponding device model can be BSIM3, BSIM4, BSIM6, or other known standard or non-standard models. It is understood that the above device models are merely exemplary, and in practical applications, models corresponding to the integrated circuit device can be selected as needed. Since MOSFET transistors are one of the most commonly used devices in integrated circuits, the integrated circuit device in this embodiment is illustrated using a MOSFET transistor as an example. However, those skilled in the art will understand that the application of this application is not limited to this.

[0038] Preferably, constructing a set of hidden states, wherein the set of hidden states includes adjustable parameters of the model, further includes: each state q in the set of hidden states n The adjustable parameter corresponds to one or a combination of the adjustable parameters, wherein the adjustable parameter is the adjustable parameter in the model.

[0039] For example, q1 = {a0}, q2 = {voff}, q3 = {a0, vth0, k2}, q4 = {voff, k2, cit}. The above adjustable parameter combinations can be configured as needed.

[0040] Preferably, the method further includes, after calculating the hidden state sequence using the Viterbi algorithm, performing deduplication processing on the hidden state sequence to obtain the predicted parameter combination.

[0041] In the technical solution of this embodiment, after predicting the parameter combination, deduplication is performed to ensure the uniqueness of each parameter in the predicted parameter combination.

[0042] Preferably, the state transition probability matrix A is: A = [a ij ] N×N , where a ij =P(i t+1 =q j |i t =q i ), i=1, 2,…,N; j=1, 2,…,N; a ij Represents the state at time t in g i State, transition to q at time t+1 j The probability of a state.

[0043] The state transition probability matrix A in this embodiment is one of the means to solidify expert experience. By solidifying expert experience and combining it with the hidden Markov algorithm, scientific prediction of parameter combinations can be achieved, automatically predicting and prompting parameter combinations, guiding users, reducing the experience threshold for users, and improving the efficiency of parameter selection by technical personnel.

[0044] Preferably, the observation probability matrix B is: B = [b j (k)] N×M , where b j (k)=P(o t =v k |i t =q j ), k=1, 2,…,M; j=1, 2,…,N; b j (k) represents the state at time t, q iState generation observation v k The probability of generation is also called the probability of emission.

[0045] The observation probability matrix B in this embodiment is one of the means to solidify expert experience. By solidifying expert experience and combining it with the hidden Markov algorithm, scientific prediction of parameter combinations can be achieved. The algorithm can automatically predict and prompt parameter combinations, guide users, reduce the experience threshold for users, and improve the efficiency of parameter selection by technical personnel.

[0046] Preferably, the method of calculating the hidden state sequence using the Viterbi algorithm based on the selected region and the obtained state transition probability matrix and observation probability matrix to obtain the predicted parameter combination further includes: determining the observation sequence O = (o1, o2, ... o3) based on the selected region. T ), where T is the number of selected regions; combining the state transition probability matrix A, the observation probability matrix B, and the number of selected regions T, the Viterbi algorithm is used to calculate k hidden state sequences that may generate observation event sequences, so as to obtain the predicted k sets of parameter combinations.

[0047] This embodiment specifically discloses a method for constructing an observation sequence O based on a user-selected region, and using the Viterbi algorithm to calculate the dimensional bit paths (i.e., k hidden state sequences) that may generate observation event sequences based on the state transition probability matrix A, the observation probability matrix B, and the number of selected regions T. The sequences are sorted according to their probability magnitudes, such as k1 = {q1, q3} = {a0, a0, vth0, k2}, k2 = {q3, q4} = {a0, vth0, k2, voff, k2, cit}. For the output k sequences, duplicate parameters within the same sequence are excluded, i.e., k1 = {a0, vth0, k2}, k2 = {a0, vth0, k2, voff, cit}. This yields the predicted k sets of parameter combinations.

[0048] Second Embodiment

[0049] Based on the same concept, this invention also provides an automatic parameter selection system for Hidden Markov Models, comprising: a selected region acquisition module, used to determine a selected region in the model in response to a selection command input by a user in a visual operation interface, and construct a set of selected regions; a parameter combination construction module, used to construct a set of hidden states, wherein the set of hidden states contains adjustable parameters of the model; a matrix acquisition module, used to acquire a state transition probability matrix and an observation probability matrix; and a prediction module, used to calculate a sequence of hidden states using the Viterbi algorithm based on the selected regions and the acquired state transition probability matrix and observation probability matrix to obtain the predicted parameter combination.

[0050] The technical solution of this embodiment is based on the Hidden Markov Model (HMM) to predict the combination of selected parameters according to the selected region. First, the expert experience is solidified in the state transition matrix and the observation probability matrix. When the user selects a region, the system will automatically predict and prompt the parameter combination based on the constructed HMM and the state transition matrix and observation probability matrix solidified by the expert experience, guide the user, reduce the experience threshold for the user, and improve the efficiency of parameter selection by technical personnel.

[0051] Third Embodiment

[0052] Based on the same concept, the present invention also provides an electronic device, characterized in that it includes: a memory for storing a processing program; and a processor, which, when executing the processing program, implements the automatic selection method for model parameters based on Hidden Markov Models as described above.

[0053] Based on the same concept, the present invention also provides a readable storage medium, characterized in that the readable storage medium stores a processing program, which, when executed by a processor, implements the automatic selection method for model parameters based on Hidden Markov as described above.

[0054] If the method for automatic parameter selection based on Hidden Markov Models is implemented as a program instruction and sold or used as an independent product, it can be stored in a computer-readable storage medium and the following steps can be performed:

[0055] S100: In response to a selection command input by the user in the visual operation interface, determine the selected region in the model and construct a set of selected regions;

[0056] S200: Construct a set of hidden states, wherein the set of hidden states contains adjustable parameters of the model;

[0057] S300: Obtain the state transition probability matrix and observation probability matrix;

[0058] S400: Based on the selected region and the obtained state transition probability matrix and observation probability matrix, the Viterbi algorithm is used to calculate the hidden state sequence to obtain the predicted parameter combination.

[0059] Based on this understanding, the technical solution of this embodiment, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in software. This computer software is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this disclosure. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0060] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific identification content executed by the system and device described above can be referred to the corresponding process in the foregoing method embodiments.

[0061] The embodiments of the present invention have been described in detail above with reference to the accompanying drawings, but the present invention is not limited to the above embodiments. Even if various changes are made to the present invention, if these changes fall within the scope of the claims of the present invention and their equivalents, they shall still fall within the protection scope of the present invention.

Claims

1. A method for automatically selecting model parameters based on Hidden Markov Models, characterized in that, Includes the following steps: In response to the selection command input by the user in the visual operation interface, the selected area in the semiconductor device model is determined and a set of selected areas is constructed; Construct a set of hidden states, wherein the set of hidden states contains adjustable parameters of the semiconductor device model; Obtain the state transition probability matrix and observation probability matrix, which are solidified from expert experience; Based on the selected region and the obtained state transition probability matrix and observation probability matrix, the Viterbi algorithm is used to calculate the hidden state sequence to obtain the predicted parameter combination.

2. The method for automatically selecting model parameters based on Hidden Markov Models according to claim 1, characterized in that, Responding to a selection command input by the user in the visual interface to determine a selected region in the model, constructing a set of selected regions further includes: Based on the representation graphic selected by the cursor in the current visualization operation interface, obtain the attribute information of the representation graphic; Obtain parameter information for all points in the marked state in the representation graph; The selected region in the model is determined based on the attribute information of the representation graph and the parameter information of all points in the marked state in the representation graph. Construct a set of selected regions V = {v1, v2, ..., v m }, m represents the number of possible observations.

3. The method for automatically selecting model parameters based on Hidden Markov Models according to claim 1, characterized in that, Constructing a set of hidden states, wherein the set of hidden states includes adjustable parameters of the model, further includes: Construct a hidden state set Q = {q1, q2, ..., q} n }, where n represents the number of possible states; The model employs one of the following models: BSIM3, BSIM4, BSIM6, BSIM-CMG, BSIM-IMG, BSIMSOI, UTSOI, HiSIM2, HiSIM_HV, PSP, GP-BJT, RPITFT, where each state q in the hidden state set... n These correspond to the adjustable parameters in the model.

4. The method for automatically selecting model parameters based on Hidden Markov Models according to claim 1 or 3, characterized in that, Constructing a set of hidden states, wherein the set of hidden states includes adjustable parameters of the model, further includes: Each state q in the hidden state set n The adjustable parameter corresponds to one or a combination of the adjustable parameters, wherein the adjustable parameter is the adjustable parameter in the model.

5. The method for automatically selecting model parameters based on Hidden Markov Models according to claim 1, characterized in that, The method further includes, after calculating the hidden state sequence using the Viterbi algorithm, performing deduplication on the hidden state sequence to obtain the predicted parameter combination.

6. The method for automatically selecting model parameters based on Hidden Markov Models according to claim 1, characterized in that, The state transition probability matrix A is: A = [a ij ] N×N , Among them, a ij =P(I t+1 =q j |I t =q i ), i=1,2,…,N; j=1,2,…,N; a ij Represents the state at time t as q i State, transition to q at time t+1 j The probability of a state.

7. The method for automatically selecting model parameters based on Hidden Markov Models according to claim 6, characterized in that, The observation probability matrix B is: B = [b j (k)] N×M , Among them, b j (k)=P(o t =v k |I t =q j ), k=1,2…,M; j=1,2,…,N; b j (k) represents the state at time t, q j State generation observation v k The probability of generation is also called the probability of emission.

8. The method for automatically selecting model parameters based on Hidden Markov Models according to claim 7, characterized in that, Based on the selected region and the obtained state transition probability matrix and observation probability matrix, the Viterbi algorithm is used to calculate the hidden state sequence to obtain the predicted parameter combination, which further includes: Based on the selected region, an observation sequence O = (o1, o2, ... o2) is determined. T ), where T is the number of selected regions; By combining the state transition probability matrix A, the observation probability matrix B, and the number of selected regions T, the Viterbi algorithm is used to calculate r hidden state sequences that may generate observation event sequences, so as to obtain r sets of predicted parameter combinations.

9. An automatic parameter selection system for models based on Hidden Markov Models, characterized in that, include: The selected region acquisition module is used to determine the selected region in the semiconductor device model in response to the selection command input by the user in the visual operation interface, and to construct a set of selected regions; A parameter combination construction module is used to construct a set of hidden states, wherein the set of hidden states contains adjustable parameters of the semiconductor device model; The matrix acquisition module is used to acquire the state transition probability matrix and observation probability matrix, which are solidified from expert experience. The prediction module is used to calculate the hidden state sequence based on the selected region and the obtained state transition probability matrix and observation probability matrix using the Viterbi algorithm to obtain the predicted parameter combination.

10. An electronic device, characterized in that, include: The memory is used to store the processing program; A processor, which, when executing the processing program, implements the method for automatic selection of model parameters based on Hidden Markov as described in any one of claims 1 to 8.

Images