# US 11,909,482 B2 · Federated learning for client-specific neural network parameter generation for wireless communication

> Agents: this data is available as JSON from the patentgrep API and MCP server — see https://patentgrep.com/llms.txt

- Number: US 11,909,482 B2
- Document ID: US11909482B2
- Status: Granted
- Kind: B2
- Application number: 17/444,028
- Filed: 2021-07-29
- Published: 2022-02-24
- Granted: 2024-02-20
- Earliest claimed: 2020-08-18
- Assignees: QUALCOMM Incorporated
- Applicants: QUALCOMM Incorporated
- Inventors: June Namgoong; Taesang Yoo; Naga Bhushan; Jay Kumar Sundararajan; Pavan Kumar Vitthaladevuni; Krishna Kiran Mukkavilli; Hwan Joon Kwon; Tingfang Ji
- Web page: https://patentgrep.com/patent/US11909482B2

## CPC

- G06N 3/088: Non-supervised learning, e.g. competitive learning
- H04B 7/0626: {Channel coefficients, e.g. channel state information [CSI]}
- G06N 3/045: Combinations of networks
- G06N 3/0455: Auto-encoder networks; Encoder-decoder networks
- G06N 3/047: Probabilistic or stochastic networks
- G06N 3/0495: Quantised networks; Sparse networks; Compressed networks
- G06N 3/098: Distributed learning, e.g. federated learning
- H04B 7/0663: {using vector or matrix manipulations}
- G06N: COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- H04B: TRANSMISSION

## Abstract

Various aspects of the present disclosure generally relate to wireless communication. In some aspects, a client may determine, using a conditioning network and based at least in part on an observed environmental vector, a set of client-specific parameters. The client may determine a latent vector using a client autoencoder and based at least in part on the set of client-specific parameters and the set of shared parameters. The client may transmit the observed environmental vector and the latent vector to a server. Numerous other aspects are provided.

## Claims

### Claim 1

A method of wireless communication performed by a client, comprising:
determining, using a conditioning network, a set of client-specific encoder parameters based at least in part on an observed environmental vector, the observed environmental vector being associated with one or more features associated with an environment of the client;
determining a latent vector using an autoencoder and based at least in part on the set of client-specific encoder parameters and an observed wireless communication vector, the observed wireless communication vector being associated with a wireless communication task; and
transmitting the observed environmental vector and the latent vector using a physical uplink control channel or a physical uplink shared channel.

### Claim 2 (depends on 1)

The method of claim 1, wherein the conditioning network is configured to receive the observed environmental vector as input and to provide a conditioning vector as output, wherein the conditioning vector comprises:
the set of client-specific encoder parameters, and
a set of client-specific decoder parameters.

### Claim 3

A method of wireless communication performed by a server, comprising:
receiving, from a client, an observed environmental vector associated with one or more features associated with an environment of the client;
receiving, from the client via a physical uplink control channel or a physical uplink shared channel, a latent vector that is based at least in part on a set of client-specific encoder parameters and an observed wireless communication vector, the observed wireless communication vector being associated with a wireless communication task;
determining the observed wireless communication vector based at least in part on the latent vector and the observed environmental vector; and
performing a wireless communication action based at least in part on determining the observed wireless communication vector.

### Claim 4

An apparatus for wireless communication at a client, comprising:
one or more memories; and
one or more processors, coupled to the one or more memories, the one or more processors configured to cause the apparatus to:
determine, using a conditioning network, a set of client-specific encoder parameters based at least in part on an observed environmental vector, the observed environmental vector being associated with one or more features associated with an environment of the client;
determine a latent vector using an autoencoder and based at least in part on the set of client-specific encoder parameters and an observed wireless communication vector, the observed wireless communication vector being associated with a wireless communication task; and
transmit the observed environmental vector and the latent vector using a physical uplink control channel or a physical uplink shared channel.

### Claim 5 (depends on 4)

The apparatus of claim 4, wherein the conditioning network is configured to receive the observed environmental vector as input and to provide a conditioning vector as output, wherein the conditioning vector comprises:
the set of client-specific encoder parameters, and
a set of client-specific decoder parameters.

### Claim 6 (depends on 5)

The apparatus of claim 5, wherein the conditioning network comprises:
a first sub-network configured to determine the set of client-specific encoder parameters; and
a second sub-network configured to determine the set of client-specific decoder parameters.

### Claim 7 (depends on 4)

The apparatus of claim 4, wherein the autoencoder comprises:
an encoder configured to receive the observed wireless communication vector as input and to provide the latent vector as output; and
a decoder configured to receive the latent vector as input and to provide the observed wireless communication vector as output.

### Claim 8 (depends on 7)

The apparatus of claim 7, wherein the encoder comprises at least one encoder layer configured to use the set of client-specific encoder parameters, and wherein, to determine the latent vector, the one or more processors are configured to cause the apparatus to:
map the observed wireless communication vector to the latent vector using the at least one encoder layer.

### Claim 9 (depends on 8)

The apparatus of claim 8, wherein, to map the observed wireless communication vector to the latent vector, the one or more processors are configured to cause the apparatus to map the observed wireless communication vector to the latent vector using a set of encoder parameters,
wherein the set of encoder parameters comprises:
the set of client-specific encoder parameters, and
a set of shared encoder parameters.

### Claim 10 (depends on 9)

The apparatus of claim 9, wherein the at least one encoder layer comprises:
a first set of layers that uses a first subset of the set of client-specific encoder parameters;
a second set of layers that uses a first subset of the set of shared encoder parameters;
and a third set of layers that uses a second subset of the set of client-specific encoder parameters and a second subset of the set of shared encoder parameters.

### Claim 11 (depends on 7)

The apparatus of claim 7, wherein the decoder comprises at least one decoder layer configured to use a set of client-specific decoder parameters, wherein the decoder is configured to determine the observed wireless communication vector by mapping the latent vector to the observed wireless communication vector using the at least one decoder layer.

### Claim 12 (depends on 11)

The apparatus of claim 11, wherein, to map the latent vector to the observed wireless communication vector, the one or more processors are configured to cause the apparatus to map the latent vector to the observed wireless communication vector using a set of decoder parameters,
wherein the set of decoder parameters comprises:
the set of client-specific decoder parameters, and
a set of shared decoder parameters.

### Claim 13 (depends on 12)

The apparatus of claim 12, wherein the at least one decoder layer comprises:
a first set of layers that uses a first subset of the set of client-specific decoder parameters;
a second set of layers that uses a first subset of the set of shared decoder parameters; and
a third set of layers that uses a second subset of the set of client-specific decoder parameters and a second subset of the set of shared decoder parameters.

### Claim 14 (depends on 4)

The apparatus of claim 4, wherein the observed environmental vector comprises one or more feature components, wherein the one or more feature components indicate:
a client vendor identifier,
a client antenna configuration,
a large scale channel characteristic,
a channel state information reference signal configuration,
an image obtained by an imaging device, or
a portion of an estimated propagation channel.

### Claim 15 (depends on 14)

The apparatus of claim 14, wherein the large scale channel characteristic indicates:
a delay spread associated with a channel,
a power delay profile associated with a channel,
a Doppler measurement associated with a channel,
a Doppler spectrum associated with a channel,
a signal to noise ratio associated with a channel,
a signal to noise plus interference ratio associated with a channel,
a reference signal received power, or
a received signal strength indicator.

### Claim 16 (depends on 4)

The apparatus of claim 4, wherein the latent vector is associated with the wireless communication task.

### Claim 17 (depends on 16)

The apparatus of claim 16, wherein the wireless communication task comprises:
a determination of channel state feedback (CSF),
a determination of positioning information associated with the client,
a determination of a modulation associated with a wireless communication, or
a determination of a waveform associated with a wireless communication.

### Claim 18 (depends on 17)

The apparatus of claim 17, wherein the wireless communication task comprises determining the CSF, and wherein the one or more processors are further configured to cause the apparatus to:
receive a channel state information (CSI) reference signal (CSI-RS);
determine CSI based at least in part on the CSI-RS; and
provide the CSI as input to the autoencoder.

### Claim 19 (depends on 18)

The apparatus of claim 18, wherein the latent vector comprises compressed channel state feedback.

### Claim 20 (depends on 4)

The apparatus of claim 4, wherein the autoencoder comprises a variational autoencoder.

### Claim 21 (depends on 4)

The apparatus of claim 4, further comprising training at least one of the conditioning network or the autoencoder, wherein training the at least one of the conditioning network or the autoencoder comprises using an unsupervised learning procedure.

### Claim 22 (depends on 4)

The apparatus of claim 4, wherein the one or more processors are further configured to cause the apparatus to collaboratively train the conditioning network and the autoencoder using a federated learning procedure.

### Claim 23 (depends on 22)

The apparatus of claim 22, wherein, to collaboratively train the conditioning network and the autoencoder, the one or more processors are configured to cause the apparatus to determine a set of shared neural network parameters that maximize a variational lower bound function corresponding to the conditioning network and the autoencoder.

### Claim 24 (depends on 23)

The apparatus of claim 23, wherein a negative variational lower bound function corresponds to a loss function.

### Claim 25 (depends on 22)

The apparatus of claim 22, wherein, to collaboratively train the conditioning network and the autoencoder, the one or more processors are configured to cause the apparatus to jointly train the conditioning network and the autoencoder.

### Claim 26 (depends on 22)

The apparatus of claim 22, wherein, to collaboratively train the conditioning network and the autoencoder, the one or more processors are configured to cause the apparatus to alternate between training the conditioning network and training the autoencoder.

### Claim 27 (depends on 26)

The apparatus of claim 26, wherein, to alternate between training the conditioning network and training the autoencoder, the one or more processors are configured to cause the apparatus to:
perform a first plurality of training iterations associated with the conditioning network according to a first training frequency; and
perform a second plurality of training iterations associated with the autoencoder according to a second training frequency that is different than the first training frequency.

### Claim 28 (depends on 4)

The apparatus of claim 4, wherein the conditioning network corresponds to a server conditioning network that is implemented at a server, and wherein the conditioning network is a client implemented copy of the server conditioning network.

### Claim 29 (depends on 4)

The apparatus of claim 4, wherein the autoencoder corresponds to a server autoencoder that is implemented at a server, and wherein the autoencoder is a client implemented copy of the server autoencoder.

### Claim 30

An apparatus for wireless communication at a server, comprising:
one or more memories; and
one or more processors, coupled to the one or more memories, the one or more processors configured to cause the apparatus to:
receive, from a client, an observed environmental vector associated with one or more features associated with an environment of the client;
receive, from the client via a physical uplink control channel or a physical uplink shared channel, a latent vector that is based at least in part on a set of client-specific encoder parameters and an observed wireless communication vector, the observed wireless communication vector being associated with a wireless communication task;
determine the observed wireless communication vector based at least in part on the latent vector and the observed environmental vector; and
perform a wireless communication action based at least in part on determining the observed wireless communication vector.