Wearable sensors have become ubiquitous thanks to a variety of health tracking features. The resulting continuous and longitudinal measurements from everyday life generate large volumes of data. However, making sense of these observations for scientific and actionable insights is non-trivial. Inspired by the empirical success of generative modeling, where large neural networks learn powerful representations from vast amounts of text, image, video, or audio data, we investigate the scaling properties of wearable sensor foundation models across compute, data, and model size. Using a dataset of up to 40 million hours of in-situ heart rate, heart rate variability, accelerometer, electrodermal activity, skin temperature, and altimeter per-minute data from over 165,000 people, we create LSM, a multimodal foundation model built on the largest wearable-signals dataset with the most extensive range of sensor modalities to date. Our results establish the scaling laws of LSM for tasks such as imputation, interpolation and extrapolation across both time and sensor modalities. Moreover, we highlight how LSM enables sample-efficient downstream learning for tasks including exercise and activity recognition.

We present SensorLM, a family of sensor-language foundation models that enable wearable sensor data understanding with natural language. Despite its pervasive nature, aligning and interpreting sensor data with language remains challenging due to the lack of paired, richly annotated sensor-text descriptions in uncurated, real-world wearable data. We introduce a hierarchical caption generation pipeline designed to capture statistical, structural, and semantic information from sensor data. This approach enabled the curation of the largest sensor-language dataset to date, comprising over 59.7 million hours of data from more than 103,000 people. Furthermore, SensorLM extends prominent multimodal pretraining architectures (e.g., CLIP, CoCa) and recovers them as specific variants within a generic architecture. Extensive experiments on real-world tasks in human activity analysis and healthcare verify the superior performance of SensorLM over state-of-the-art in zero-shot recognition, few-shot learning, and cross-modal retrieval. SensorLM also demonstrates intriguing capabilities including scaling behaviors, label efficiency, sensor captioning, and zero-shot generalization to unseen tasks. Code is available at https://github.com/Google-Health/consumer-health-research/tree/main/sensorlm.

Based on the recent compute-and-forward technique, a novel communication strategy is proposed under which functions of the channel state information are forwarded along the network. Those functions are chosen such that on the one hand, they can be efficiently forwarded, and on the other hand, they are maximally useful to the final decoder of the message. It is illustrated that there is generally a tension between these two requirements. The strategy is shown to perform well for certain classes of multilayer networks where channel state information is acquired locally at each receiver. For example, for a two-stage Gaussian relay network with local channel state information, it is shown that the proposed strategy performs optimally in a scaling-law sense, as the number of relays increases.

Linear receivers are often used to reduce the implementation complexity of multiple-antenna systems. In a traditional linear receiver architecture, the receive antennas are used to separate out the codewords sent by each transmit antenna, which can then be decoded individually. Although easy to implement, this approach can be highly suboptimal when the channel matrix is near singular. This paper develops a new linear receiver architecture that uses the receive antennas to create an effective channel matrix with integer-valued entries. Rather than attempting to recover transmitted codewords directly, the decoder recovers integer combinations of the codewords according to the entries of the effective channel matrix. The codewords are all generated using the same linear code, which guarantees that these integer combinations are themselves codewords. Provided that the effective channel is full rank, these integer combinations can then be digitally solved for the original codewords. This paper focuses on the special case where there is no coding across transmit antennas and no channel state information at the transmitter(s), which corresponds either to a multiuser uplink scenario or to single-user V-BLAST encoding. In this setting, the proposed integer-forcing linear receiver significantly outperforms conventional linear architectures such as the zero forcing and linear minimum mean-squared error receiver. In the high signal-to-noise ratio regime, the proposed receiver attains the optimal diversity-multiplexing tradeoff for the standard multiple-input multiple-output (MIMO) channel with no coding across transmit antennas. It is further shown that in an extended MIMO model with interference, the integer-forcing linear receiver achieves the optimal generalized degrees of freedom.

In linear function computation, multiple source nodes communicate across a relay network to a single destination whose goal is to recover linear functions of the original source data. When the relay network is a linear deterministic network, a duality relation is established between function computation and broadcast with common messages. Using this relation, a compact sufficient condition is found describing those cases where the cut-set bound is tight. These insights are used to develop results for the case where the relay network contains Gaussian multiple-access channels. The proposed scheme decouples the physical and network layers. Using lattice codes for both source quantization and computation in the physical layer, the original Gaussian sources are converted into discrete sources and the Gaussian network into a linear deterministic network. Network codes for computing functions of discrete sources across the deterministic network are then found by applying the duality relation. The distortion for computing the sum of an arbitrary number of independent Gaussian sources over the Gaussian network is proven to be within a constant factor of the optimal performance. Furthermore, the constant factor results are extended to include asymmetric functions for the case of two sources.

The Compute-and-Forward approach has been proven to be very beneficial for communication over Gaussian networks. While the theoretical results are promising, it is still not completely understood how to best apply this scheme in practice. The objective of this work is to provide a low complexity scheme suitable for Compute-and-Forward. The scheme is based on utilizing linear codes over ℤqwhere q is not restricted to be prime and allows to achieve high transmission rates following Ungerboeck's set partitioning principle.

We show that the recently proposed integer-forcing linear receiver provides an attractive approach to the problem of mitigating external interference in MIMO channels. The integer-forcing receiver proceeds by first decoding a set of full rank integer linear combinations of the data streams. The resulting full rank equations are then inverted to find the original data. By selecting equation coefficients in a direction that depends on both the interference space and the channel matrix, the impact of external interference can be effectively reduced. We show that this technique attains a non-trivial gain over traditional linear receivers. Furthermore, the integer-forcing linear receiver achieves the same generalized degrees of freedom for the M×M MIMO channel with K dimensional external interference as the joint decoder.

Linear receivers are often used to reduce the implementation complexity of multiple antenna systems. In a traditional linear receiver architecture, the receive antennas are used to separate out the codewords sent by each transmit antenna, which can then be decoded individually. Although easy to implement, this approach can be highly sub-optimal when the channel matrix is near singular. In this paper, we develop a new linear architecture that uses the receive antennas to create an effective channel matrix with integer-valued entries. Instead of attempting to recover a transmitted codeword directly, each decoder recovers a different integer combination of the codewords according to the effective channel matrix. If the effective channel is full rank, these linear equations can be digitally solved for the original codewords. By allowing the receiver to equalize the channel to any matrix with integer entries, this scheme can outperform traditional linear architectures such as decorrelators and MMSE receivers while maintaining a similar complexity. Furthermore, in the case where each transmit antenna encodes an independent data stream, the proposed receiver attains the optimal diversity multiplexing tradeoff.

We propose a new framework for MIMO decoding based on a recently developed technique for reliably conveying linear equations over wireless channels. Each transmit antenna sends an independent data stream using the same linear code. As a result, any integer combination of the codewords is itself a codeword. Each receive antenna observes a random complex-valued combination of the codewords according to the fading coefficients. We use a linear pre-processing step at the receiver to transform the effective channel into a (full-rank) integer matrix. A single stream decoder is then used to recover integer combinations of the codewords. These equations of codewords are then translated into equations of the transmitted data streams over a finite field which can be easily solved for the original data. We examine the performance of our scheme in terms of the probability of outage and show that significant gains are possible over standard linear architectures for both i.i.d. and correlated Rayleigh fading.

In many network communication scenarios, a relay in the network may only need to recover and retransmit an equation of the transmitted messages. In previous work, it has been shown that if each transmitter employs the same lattice code, the interference structure of the channel can be exploited to recover an equation much more efficiently than possible with standard multiple-access strategies. Here, we generalize this compute-and-forward framework to the multiple antenna setting. Our results show that it is often beneficial to use extra antennas at the receiver to rotate the channel coefficients towards the nearest integer vector instead of separating out the transmitted signals. We also demonstrate that in contrast to classical strategies, the multiplexing gain of compute-and-forward increases if the transmitters have channel state information. Finally, we apply our scheme to the two way relay network and observe performance gains over traditional strategies.

In large relay networks, the assumption of full and perfect channel knowledge at the destination is optimistic in practice. The fading coefficients are typically measured at the relays but not directly known at the destination. Traditionally, each fading coefficient is individually forwarded to the destination. However, it is often sufficient for the decoder to know only a function of the various channel states rather than the full information. We develop a general framework for forwarding channel state information in relay systems with local channel knowledge. We apply our framework to several networks and find that functional forwarding of channel state information can be attained much more efficiently than full forwarding.