Patent
Atlas literature
Patent
US 12,561,498 B2Patent drawings and their descriptions. Click a drawing to enlarge it.
FIG. 1 shows a schematic diagram illustrating an over- view of the multi-agent RL framework for automated design of RF circuits, according to embodiments of …
FIG. 2 show a schematic of N-cell non-uniform distrib- uted power amplifier, according to some embodiments of the present disclosure;
FIG. 3 shows a multi-agent RL model embedding based 40 on actor-critic algorithm, according to some embodiments of the present disclosure;
FIG. 4 Shows a policy network encoding for individual agent, according to some embodiments of the present dis- closure; 45
FIG. 5 shows a table of specification (Scattering param- eters or S-parameters) for 3-cell non-uniform distributed power amplifier, according to some embodiments …
FIG. 6 shows a histogram of how much specifications are 50 met after sufficient training in MARL, according to some embodiments of the present disclosure
FIG. 7 shows the comparison of training results with single-agent RL and multi-agent RL for 3-cell NDPA, according to some embodiments of the present …
FIG. 8 shows a schematic of a system configured with processor, memory coupled with a storage storing computer- implemented methods, according to some …
Claims define the patent's legal scope. Independent claims stand alone; dependent claims (nested) narrow them. Click a claim to expand its dependents.
A computer-implemented method for training a multiagent reinforcement learning (RL) network generating device parameters of circuits, wherein the method uses a processor coupled with a memory storing instructions imple-menting the method, wherein the instructions, when executed by the processor, carry out at steps of the method, comprising: acquiring inputs with respect to a desired circuit specifi-cation of a circuit, device parameters of the circuit, a topology of the circuit, a final state corresponding to a maximum step, wherein the desired circuit specifica-tion is represented by a gain, bandwidth, phase margin, power consumption, output power and power effi-ciency, or part of combination thereof; decomposing the circuit into a plurality of amplifying units, each amplifying unit is modeled as an agent in the multi-agent RL network, wherein each agent of the multi-agent RL network is configured to perform steps of: transmitting an action selected from a set of actions to an environment module that includes a netlist of the circuit; updating the device parameters of the circuit with respect to the desired circuit specification according to the selected action using a data processor of the environ-ment module, wherein the action changes each of the device parameters by a minimum unit value of each device parameter; obtaining a current circuit specification of the circuit by simulating the netlist of the circuit based on the updated device parameters using a circuit simulator of the environment module; acquiring a reward from the environment module, wherein the reward is computed based on a difference between the current circuit specification and the desired circuit specification, and a policy of each agent is B₂ updated based on the reward and a state value of the circuit that is estimated by a shared graphical attention network based on the desired circuit specification and features extracted from all agents of the multi-agent RL network, wherein the steps of the transmitting, updat-ing, obtaining and acquiring are continued until the reward reaches to a threshold value or a number of the steps reach a preset value; and storing the updated device parameters into the memory.
The method of claim 1, wherein the circuit is modeled by a graph G(V, E), wherein each node V is represented by a device, wherein an edge E represents a connection between devices.
The method of claim 1, wherein the power supply (VP), ground (VGND), and other DC bias voltages in the topology of the circuit are represented as extra nodes V.
The method of claim 1, wherein the reward is calcu-lated by a weighted sum of the desired specification by assigning different value of the weighting factor.
The method of claim 1, wherein the multi-agent RL network includes a graph neural network (GNN) and a fully connected neural network (FCNN).
A computer-implemented method for generating device parameters of circuits using a pretrained multi-agent rein-forcement learning (RL) network, wherein the method uses a processor coupled with a memory storing instructions implementing the method, wherein the instructions, when executed by the processor, carry out at steps of the method, comprising: acquiring, via an interface, inputs with respect to a desired specification of a circuit, device parameters, a topology of the circuit, wherein the circuit corresponds a distrib-uted circuit including a plurality of amplifying units, each amplifying unit is modeled as an agent in the pretrained multi-agent RL network; providing the inputs to the pretrained multi-agent RL network, wherein each of the desired circuit specifica-tion is represented by gain, bandwidth, phase margin, power consumption, output power and power effi-ciency, wherein a policy of each agent is based on a reward computed based on a difference between current circuit specification and the desired circuit specification and a state value of the circuit that is estimated by a shared graphical attention network based on the desired circuit specification and features extracted from all agents of the multi-agent RL network; and generating a circuit represented by a graph modeling the topology of the circuit and the device parameters of the circuit, updated device parameters of the circuit. ∗ ∗ ∗ ∗ ∗
Layer stacks claimed or described, ordered top of device to substrate.
GaN distributed RF power amplifier
No layer stack recorded.
Patents and literature cited by this patent (applicant and examiner references).
Cited patents · 2
Cited non-patent literature · 7
Patent
Atlas literature
Patent
US 12,561,498 B2Patent drawings and their descriptions. Click a drawing to enlarge it.
FIG. 1 shows a schematic diagram illustrating an over- view of the multi-agent RL framework for automated design of RF circuits, according to embodiments of …
FIG. 2 show a schematic of N-cell non-uniform distrib- uted power amplifier, according to some embodiments of the present disclosure;
FIG. 3 shows a multi-agent RL model embedding based 40 on actor-critic algorithm, according to some embodiments of the present disclosure;
FIG. 4 Shows a policy network encoding for individual agent, according to some embodiments of the present dis- closure; 45
FIG. 5 shows a table of specification (Scattering param- eters or S-parameters) for 3-cell non-uniform distributed power amplifier, according to some embodiments …
FIG. 6 shows a histogram of how much specifications are 50 met after sufficient training in MARL, according to some embodiments of the present disclosure
FIG. 7 shows the comparison of training results with single-agent RL and multi-agent RL for 3-cell NDPA, according to some embodiments of the present …
FIG. 8 shows a schematic of a system configured with processor, memory coupled with a storage storing computer- implemented methods, according to some …
Claims define the patent's legal scope. Independent claims stand alone; dependent claims (nested) narrow them. Click a claim to expand its dependents.
A computer-implemented method for training a multiagent reinforcement learning (RL) network generating device parameters of circuits, wherein the method uses a processor coupled with a memory storing instructions imple-menting the method, wherein the instructions, when executed by the processor, carry out at steps of the method, comprising: acquiring inputs with respect to a desired circuit specifi-cation of a circuit, device parameters of the circuit, a topology of the circuit, a final state corresponding to a maximum step, wherein the desired circuit specifica-tion is represented by a gain, bandwidth, phase margin, power consumption, output power and power effi-ciency, or part of combination thereof; decomposing the circuit into a plurality of amplifying units, each amplifying unit is modeled as an agent in the multi-agent RL network, wherein each agent of the multi-agent RL network is configured to perform steps of: transmitting an action selected from a set of actions to an environment module that includes a netlist of the circuit; updating the device parameters of the circuit with respect to the desired circuit specification according to the selected action using a data processor of the environ-ment module, wherein the action changes each of the device parameters by a minimum unit value of each device parameter; obtaining a current circuit specification of the circuit by simulating the netlist of the circuit based on the updated device parameters using a circuit simulator of the environment module; acquiring a reward from the environment module, wherein the reward is computed based on a difference between the current circuit specification and the desired circuit specification, and a policy of each agent is B₂ updated based on the reward and a state value of the circuit that is estimated by a shared graphical attention network based on the desired circuit specification and features extracted from all agents of the multi-agent RL network, wherein the steps of the transmitting, updat-ing, obtaining and acquiring are continued until the reward reaches to a threshold value or a number of the steps reach a preset value; and storing the updated device parameters into the memory.
The method of claim 1, wherein the circuit is modeled by a graph G(V, E), wherein each node V is represented by a device, wherein an edge E represents a connection between devices.
The method of claim 1, wherein the power supply (VP), ground (VGND), and other DC bias voltages in the topology of the circuit are represented as extra nodes V.
The method of claim 1, wherein the reward is calcu-lated by a weighted sum of the desired specification by assigning different value of the weighting factor.
The method of claim 1, wherein the multi-agent RL network includes a graph neural network (GNN) and a fully connected neural network (FCNN).
A computer-implemented method for generating device parameters of circuits using a pretrained multi-agent rein-forcement learning (RL) network, wherein the method uses a processor coupled with a memory storing instructions implementing the method, wherein the instructions, when executed by the processor, carry out at steps of the method, comprising: acquiring, via an interface, inputs with respect to a desired specification of a circuit, device parameters, a topology of the circuit, wherein the circuit corresponds a distrib-uted circuit including a plurality of amplifying units, each amplifying unit is modeled as an agent in the pretrained multi-agent RL network; providing the inputs to the pretrained multi-agent RL network, wherein each of the desired circuit specifica-tion is represented by gain, bandwidth, phase margin, power consumption, output power and power effi-ciency, wherein a policy of each agent is based on a reward computed based on a difference between current circuit specification and the desired circuit specification and a state value of the circuit that is estimated by a shared graphical attention network based on the desired circuit specification and features extracted from all agents of the multi-agent RL network; and generating a circuit represented by a graph modeling the topology of the circuit and the device parameters of the circuit, updated device parameters of the circuit. ∗ ∗ ∗ ∗ ∗
Layer stacks claimed or described, ordered top of device to substrate.
GaN distributed RF power amplifier
No layer stack recorded.
Patents and literature cited by this patent (applicant and examiner references).
Cited patents · 2
Cited non-patent literature · 7
Patent
Atlas literature
Patent
US 12,561,498 B2Patent drawings and their descriptions. Click a drawing to enlarge it.
FIG. 1 shows a schematic diagram illustrating an over- view of the multi-agent RL framework for automated design of RF circuits, according to embodiments of …
FIG. 2 show a schematic of N-cell non-uniform distrib- uted power amplifier, according to some embodiments of the present disclosure;
FIG. 3 shows a multi-agent RL model embedding based 40 on actor-critic algorithm, according to some embodiments of the present disclosure;
FIG. 4 Shows a policy network encoding for individual agent, according to some embodiments of the present dis- closure; 45
FIG. 5 shows a table of specification (Scattering param- eters or S-parameters) for 3-cell non-uniform distributed power amplifier, according to some embodiments …
FIG. 6 shows a histogram of how much specifications are 50 met after sufficient training in MARL, according to some embodiments of the present disclosure
FIG. 7 shows the comparison of training results with single-agent RL and multi-agent RL for 3-cell NDPA, according to some embodiments of the present …
FIG. 8 shows a schematic of a system configured with processor, memory coupled with a storage storing computer- implemented methods, according to some …
Claims define the patent's legal scope. Independent claims stand alone; dependent claims (nested) narrow them. Click a claim to expand its dependents.
A computer-implemented method for training a multiagent reinforcement learning (RL) network generating device parameters of circuits, wherein the method uses a processor coupled with a memory storing instructions imple-menting the method, wherein the instructions, when executed by the processor, carry out at steps of the method, comprising: acquiring inputs with respect to a desired circuit specifi-cation of a circuit, device parameters of the circuit, a topology of the circuit, a final state corresponding to a maximum step, wherein the desired circuit specifica-tion is represented by a gain, bandwidth, phase margin, power consumption, output power and power effi-ciency, or part of combination thereof; decomposing the circuit into a plurality of amplifying units, each amplifying unit is modeled as an agent in the multi-agent RL network, wherein each agent of the multi-agent RL network is configured to perform steps of: transmitting an action selected from a set of actions to an environment module that includes a netlist of the circuit; updating the device parameters of the circuit with respect to the desired circuit specification according to the selected action using a data processor of the environ-ment module, wherein the action changes each of the device parameters by a minimum unit value of each device parameter; obtaining a current circuit specification of the circuit by simulating the netlist of the circuit based on the updated device parameters using a circuit simulator of the environment module; acquiring a reward from the environment module, wherein the reward is computed based on a difference between the current circuit specification and the desired circuit specification, and a policy of each agent is B₂ updated based on the reward and a state value of the circuit that is estimated by a shared graphical attention network based on the desired circuit specification and features extracted from all agents of the multi-agent RL network, wherein the steps of the transmitting, updat-ing, obtaining and acquiring are continued until the reward reaches to a threshold value or a number of the steps reach a preset value; and storing the updated device parameters into the memory.
The method of claim 1, wherein the circuit is modeled by a graph G(V, E), wherein each node V is represented by a device, wherein an edge E represents a connection between devices.
The method of claim 1, wherein the power supply (VP), ground (VGND), and other DC bias voltages in the topology of the circuit are represented as extra nodes V.
The method of claim 1, wherein the reward is calcu-lated by a weighted sum of the desired specification by assigning different value of the weighting factor.
The method of claim 1, wherein the multi-agent RL network includes a graph neural network (GNN) and a fully connected neural network (FCNN).
A computer-implemented method for generating device parameters of circuits using a pretrained multi-agent rein-forcement learning (RL) network, wherein the method uses a processor coupled with a memory storing instructions implementing the method, wherein the instructions, when executed by the processor, carry out at steps of the method, comprising: acquiring, via an interface, inputs with respect to a desired specification of a circuit, device parameters, a topology of the circuit, wherein the circuit corresponds a distrib-uted circuit including a plurality of amplifying units, each amplifying unit is modeled as an agent in the pretrained multi-agent RL network; providing the inputs to the pretrained multi-agent RL network, wherein each of the desired circuit specifica-tion is represented by gain, bandwidth, phase margin, power consumption, output power and power effi-ciency, wherein a policy of each agent is based on a reward computed based on a difference between current circuit specification and the desired circuit specification and a state value of the circuit that is estimated by a shared graphical attention network based on the desired circuit specification and features extracted from all agents of the multi-agent RL network; and generating a circuit represented by a graph modeling the topology of the circuit and the device parameters of the circuit, updated device parameters of the circuit. ∗ ∗ ∗ ∗ ∗
Layer stacks claimed or described, ordered top of device to substrate.
GaN distributed RF power amplifier
No layer stack recorded.
Patents and literature cited by this patent (applicant and examiner references).
Cited patents · 2
Cited non-patent literature · 7
Patent
Atlas literature
Patent
US 12,561,498 B2Patent drawings and their descriptions. Click a drawing to enlarge it.
FIG. 1 shows a schematic diagram illustrating an over- view of the multi-agent RL framework for automated design of RF circuits, according to embodiments of …
FIG. 2 show a schematic of N-cell non-uniform distrib- uted power amplifier, according to some embodiments of the present disclosure;
FIG. 3 shows a multi-agent RL model embedding based 40 on actor-critic algorithm, according to some embodiments of the present disclosure;
FIG. 4 Shows a policy network encoding for individual agent, according to some embodiments of the present dis- closure; 45
FIG. 5 shows a table of specification (Scattering param- eters or S-parameters) for 3-cell non-uniform distributed power amplifier, according to some embodiments …
FIG. 6 shows a histogram of how much specifications are 50 met after sufficient training in MARL, according to some embodiments of the present disclosure
FIG. 7 shows the comparison of training results with single-agent RL and multi-agent RL for 3-cell NDPA, according to some embodiments of the present …
FIG. 8 shows a schematic of a system configured with processor, memory coupled with a storage storing computer- implemented methods, according to some …
Claims define the patent's legal scope. Independent claims stand alone; dependent claims (nested) narrow them. Click a claim to expand its dependents.
A computer-implemented method for training a multiagent reinforcement learning (RL) network generating device parameters of circuits, wherein the method uses a processor coupled with a memory storing instructions imple-menting the method, wherein the instructions, when executed by the processor, carry out at steps of the method, comprising: acquiring inputs with respect to a desired circuit specifi-cation of a circuit, device parameters of the circuit, a topology of the circuit, a final state corresponding to a maximum step, wherein the desired circuit specifica-tion is represented by a gain, bandwidth, phase margin, power consumption, output power and power effi-ciency, or part of combination thereof; decomposing the circuit into a plurality of amplifying units, each amplifying unit is modeled as an agent in the multi-agent RL network, wherein each agent of the multi-agent RL network is configured to perform steps of: transmitting an action selected from a set of actions to an environment module that includes a netlist of the circuit; updating the device parameters of the circuit with respect to the desired circuit specification according to the selected action using a data processor of the environ-ment module, wherein the action changes each of the device parameters by a minimum unit value of each device parameter; obtaining a current circuit specification of the circuit by simulating the netlist of the circuit based on the updated device parameters using a circuit simulator of the environment module; acquiring a reward from the environment module, wherein the reward is computed based on a difference between the current circuit specification and the desired circuit specification, and a policy of each agent is B₂ updated based on the reward and a state value of the circuit that is estimated by a shared graphical attention network based on the desired circuit specification and features extracted from all agents of the multi-agent RL network, wherein the steps of the transmitting, updat-ing, obtaining and acquiring are continued until the reward reaches to a threshold value or a number of the steps reach a preset value; and storing the updated device parameters into the memory.
The method of claim 1, wherein the circuit is modeled by a graph G(V, E), wherein each node V is represented by a device, wherein an edge E represents a connection between devices.
The method of claim 1, wherein the power supply (VP), ground (VGND), and other DC bias voltages in the topology of the circuit are represented as extra nodes V.
The method of claim 1, wherein the reward is calcu-lated by a weighted sum of the desired specification by assigning different value of the weighting factor.
The method of claim 1, wherein the multi-agent RL network includes a graph neural network (GNN) and a fully connected neural network (FCNN).
A computer-implemented method for generating device parameters of circuits using a pretrained multi-agent rein-forcement learning (RL) network, wherein the method uses a processor coupled with a memory storing instructions implementing the method, wherein the instructions, when executed by the processor, carry out at steps of the method, comprising: acquiring, via an interface, inputs with respect to a desired specification of a circuit, device parameters, a topology of the circuit, wherein the circuit corresponds a distrib-uted circuit including a plurality of amplifying units, each amplifying unit is modeled as an agent in the pretrained multi-agent RL network; providing the inputs to the pretrained multi-agent RL network, wherein each of the desired circuit specifica-tion is represented by gain, bandwidth, phase margin, power consumption, output power and power effi-ciency, wherein a policy of each agent is based on a reward computed based on a difference between current circuit specification and the desired circuit specification and a state value of the circuit that is estimated by a shared graphical attention network based on the desired circuit specification and features extracted from all agents of the multi-agent RL network; and generating a circuit represented by a graph modeling the topology of the circuit and the device parameters of the circuit, updated device parameters of the circuit. ∗ ∗ ∗ ∗ ∗
Layer stacks claimed or described, ordered top of device to substrate.
GaN distributed RF power amplifier
No layer stack recorded.
Patents and literature cited by this patent (applicant and examiner references).
Cited patents · 2
Cited non-patent literature · 7
