Search

Compilers are complex, and significant effort has been expended on testing them. Techniques such as random program generation and differential testing have proved highly effective and have uncovered thousands of bugs in production compilers. The majority of effort has been expended on validating that a compiler produces correct code for a given input, while less attention has been paid to ensuring that the compiler produces performant code. In this work we adapt differential testing to the task of identifying missed optimization opportunities in compilers. We develop a novel testing approach which combines large language models (LLMs) with a series of differential testing strategies and use them to find missing code size optimizations in C / C++ compilers. The advantage of our approach is its simplicity. We offload the complex task of generating random code to an off-the-shelf LLM, and use heuristics and analyses to identify anomalous compiler behavior. Our approach requires fewer than 150 lines of code to implement. This simplicity makes it extensible. By simply changing the target compiler and initial LLM prompt we port the approach from C / C++ to Rust and Swift, finding bugs in both. To date we have reported 24 confirmed bugs in production compilers, and conclude that LLM-assisted testing is a promising avenue for detecting optimization bugs in real world compilers., Comment: Accepted to appear in The International Conference on Compiler Construction (CC) 2025

This paper introduces Data-enabled Predictive Control Hyperparameter Tuning via Differentiable Optimization (DeePC-Hunt), a backpropagation-based method for automatic hyperparameter tuning of the DeePC algorithm. The necessity for such a method arises from the importance of hyperparameter selection to achieve satisfactory closed-loop DeePC performance. The standard methods for hyperparameter selection are to either optimize the open-loop performance, or use manual guess-and-check. Optimizing the open-loop performance can result in unacceptable closed-loop behavior, while manual guess-and-check can pose safety challenges. DeePC-Hunt provides an alternative method for hyperparameter tuning which uses an approximate model of the system dynamics and backpropagation to directly optimize hyperparameters for the closed-loop DeePC performance. Numerical simulations demonstrate the effectiveness of DeePC in combination with DeePC-Hunt in a complex stabilization task for a nonlinear system and its superiority over model-based control strategies in terms of robustness to model misspecifications.

Job automation can undermine economic security for workers in general, and older workers, in particular. In this respect, consistently updating one's knowledge and skills is essential for being competitive in a technology-driven labor market. Older workers with lower adult literacy skills experience difficulties with continuous education and skills development, which can contribute to their economic uncertainty. Therefore, this study aimed to examine the correlation between adult literacy skills and job automation risk. We analyzed a nationally representative sample of middle-aged and older workers (age 50+ years; n = 1,880) from the 2012/2014/2017 Program for the International Assessment of Adult Competencies (PIAAC) dataset. The survey-weighted linear regression results showed that lower job automation risks (0%-100%) were a function of higher adult literacy skills (score 0-500 points: b = -0.052, p < 0.05), after adjusting for relevant covariates (R-squared = 0.19). [This is the online first version of an article published in "New Horizons in Adult Education and Human Resource Development."]

We address the problem of client participation in federated learning, where traditional methods typically rely on a random selection of a small subset of clients for each training round. In contrast, we propose FedBack, a deterministic approach that leverages control-theoretic principles to manage client participation in ADMM-based federated learning. FedBack models client participation as a discrete-time dynamical system and employs an integral feedback controller to adjust each client's participation rate individually, based on the client's optimization dynamics. We provide global convergence guarantees for our approach by building on the recent federated learning research. Numerical experiments on federated image classification demonstrate that FedBack achieves up to 50\% improvement in communication and computational efficiency over algorithms that rely on a random selection of clients.

Portfolio optimization is a ubiquitous problem in financial mathematics that relies on accurate estimates of covariance matrices for asset returns. However, estimates of pairwise covariance could be better and calculating time-sensitive optimal portfolios is energy-intensive for digital computers. We present an energy-efficient, fast, and fully analog pipeline for solving portfolio optimization problems that overcomes these limitations. The analog paradigm leverages the fundamental principles of physics to recover accurate optimal portfolios in a two-step process. Firstly, we utilize equilibrium propagation, an analog alternative to backpropagation, to train linear autoencoder neural networks to calculate low-rank covariance matrices. Then, analog continuous Hopfield networks output the minimum variance portfolio for a given desired expected return. The entire efficient frontier may then be recovered, and an optimal portfolio selected based on risk appetite., Comment: 7 pages, 4 figures, accepted to NeurlPS 2024

Predictive maintenance has been used to optimize system repairs in the industrial, medical, and financial domains. This technique relies on the consistent ability to detect and predict anomalies in critical systems. AI models have been trained to detect system faults, improving predictive maintenance efficiency. Typically there is a lack of fault data to train these models, due to organizations working to keep fault occurrences and down time to a minimum. For newly installed systems, no fault data exists since they have yet to fail. By using diffusion models for synthetic data generation, the complex training datasets for these predictive models can be supplemented with high level synthetic fault data to improve their performance in anomaly detection. By learning the relationship between healthy and faulty data in similar systems, a diffusion model can attempt to apply that relationship to healthy data of a newly installed system that has no fault data. The diffusion model would then be able to generate useful fault data for the new system, and enable predictive models to be trained for predictive maintenance. The following paper demonstrates a system for generating useful, multivariate synthetic data for predictive maintenance, and how it can be applied to systems that have yet to fail.

This paper presents the Complex Vector Gain-Based Annealer (CoVeGA), an analog computing platform designed to overcome energy barriers in XY Hamiltonians through a higher-dimensional representation. Traditional gain-based solvers utilizing optical or photonic hardware typically represent each XY spin with a single complex field. These solvers often struggle with large energy barriers in complex landscapes, leading to relaxation into excited states. CoVeGA addresses these limitations by employing two complex fields to represent each XY spin and dynamically evolving the energy landscape through time-dependent annealing. Operating in a higher-dimensional space, CoVeGA bridges energy barriers in this expanded space during the continuous phase evolution, thus avoiding entrapment in local minima. We introduce several graph structures that pose challenges for XY minimization and use them to benchmark CoVeGA against single-dimension XY solvers, highlighting the benefits of higher-dimensional operation., Comment: 10 pages, 8 figures

For data-constrained, complex and dynamic industrial environments, there is a critical need for transferable and multimodal methodologies to enhance anomaly detection and therefore, prevent costs associated with system failures. Typically, traditional PdM approaches are not transferable or multimodal. This work examines the use of Large Language Models (LLMs) for anomaly detection in complex and dynamic manufacturing systems. The research aims to improve the transferability of anomaly detection models by leveraging Large Language Models (LLMs) and seeks to validate the enhanced effectiveness of the proposed approach in data-sparse industrial applications. The research also seeks to enable more collaborative decision-making between the model and plant operators by allowing for the enriching of input series data with semantics. Additionally, the research aims to address the issue of concept drift in dynamic industrial settings by integrating an adaptability mechanism. The literature review examines the latest developments in LLM time series tasks alongside associated adaptive anomaly detection methods to establish a robust theoretical framework for the proposed architecture. This paper presents a novel model framework (AAD-LLM) that doesn't require any training or finetuning on the dataset it is applied to and is multimodal. Results suggest that anomaly detection can be converted into a "language" task to deliver effective, context-aware detection in data-constrained industrial applications. This work, therefore, contributes significantly to advancements in anomaly detection methodologies.

There has been a surge of interest in leveraging speech as a marker of health for a wide spectrum of conditions. The underlying premise is that any neurological, mental, or physical deficits that impact speech production can be objectively assessed via automated analysis of speech. Recent advances in speech-based Artificial Intelligence (AI) models for diagnosing and tracking mental health, cognitive, and motor disorders often use supervised learning, similar to mainstream speech technologies like recognition and verification. However, clinical speech AI has distinct challenges, including the need for specific elicitation tasks, small available datasets, diverse speech representations, and uncertain diagnostic labels. As a result, application of the standard supervised learning paradigm may lead to models that perform well in controlled settings but fail to generalize in real-world clinical deployments. With translation into real-world clinical scenarios in mind, this tutorial paper provides an overview of the key components required for robust development of clinical speech AI. Specifically, this paper will cover the design of speech elicitation tasks and protocols most appropriate for different clinical conditions, collection of data and verification of hardware, development and validation of speech representations designed to measure clinical constructs of interest, development of reliable and robust clinical prediction models, and ethical and participant considerations for clinical speech AI. The goal is to provide comprehensive guidance on building models whose inputs and outputs link to the more interpretable and clinically meaningful aspects of speech, that can be interrogated and clinically validated on clinical datasets, and that adhere to ethical, privacy, and security considerations by design., Comment: 76 pages, 24 figures

Ising machines are an emerging class of hardware that promises ultrafast and energy-efficient solutions to NP-hard combinatorial optimization problems. Spatial photonic Ising machines (SPIMs) exploit optical computing in free space to accelerate the computation, showcasing parallelism, scalability, and low power consumption. However, current SPIMs can implement only a restricted class of problems. This partial programmability is a critical limitation that hampers their benchmark. Achieving full programmability of the device while preserving its scalability is an open challenge. Here, we report a fully programmable SPIM achieved through a novel operation method based on the division of the focal plane. In our scheme, a general Ising problem is decomposed into a set of Mattis Hamiltonians, whose energies are simultaneously computed optically by measuring the intensity on different regions of the camera sensor. Exploiting this concept, we experimentally demonstrate the computation with high success probability of ground-state solutions of up to 32-spin Ising models on unweighted maximum cut graphs with and without ferromagnetic bias. Simulations of the hardware prove a favorable scaling of the accuracy with the number of spins. Our fully programmable SPIM enables the implementation of many quadratic unconstrained binary optimization problems, further establishing SPIMs as a leading paradigm in non von Neumann hardware.

Tools for rewriting, refactoring and optimizing code should be fast and correct. Large language models (LLMs), by their nature, possess neither of these qualities. Yet, there remains tremendous opportunity in using LLMs to improve code. We explore the use of LLMs not to transform code, but to code transforms. We propose a chain-of-thought approach to synthesizing code transformations from a small number of input/output code examples that incorporates execution and feedback. Unlike the direct rewrite approach, LLM-generated transformations are easy to inspect, debug, and validate. The logic of the rewrite is explicitly coded and easy to adapt. The compute required to run code transformations is minute compared to that of LLM rewriting. We test our approach on 16 Python code transformations and find that LLM- generated transforms are perfectly precise for 7 of them and less imprecise than direct LLM rewriting on the others. We hope to encourage further research to improving the precision of LLM code rewriting.

Low-density parity-check (LDPC) codes form part of the IRIG-106 standard and have been successfully deployed for the Telemetry Group version of shaped-offset quadrature phase shift keying (SOQPSK-TG) modulation. Recently, LDPC code solutions have been proposed and optimized for continuous phase modulations (CPMs), including the pulse code modulation/frequency modulation (PCM/FM) and the multi-h CPM developed by the Advanced Range TeleMetry program (ARTM CPM). These codes were shown to perform around one dB from the respective channel capacities of these modulations. In this paper, we consider the effect of random puncturing of these LDPC codes to further improve spectrum efficiency. We present numerical simulation results that affirm the robust decoding performance promised by LDPC codes designed for ARTM CPM., Comment: Accepted for inclusion in the 2024 International Telemetry Conference

High-resolution millimetre-imaging of protoplanetary discs has revealed many containing rings and gaps. These rings can contain large quantities of dust, often in excess of 10M$_\oplus$, providing prime sites for efficient and rapid planet formation. Rapid planet formation will produce high accretion luminosities, heating the surrounding disc. We investigate the importance of a planetary embryo's accretion luminosity by simulating the dynamics of the gas and dust in a dust ring, accounting for the energy liberated as a resident planetary embryo accretes. The resulting heating alters the flow structure near the planet, increasing the accretion rate of large, millimetre-to-centimetre-sized dust grains. We show how this process varies with the mass of dust in the ring and the local background gas temperature, demonstrating that the thermal feedback always acts to increase the planet's mass. This increase in planet mass is driven primarily by the formation of vortices, created by a baroclinic instability once the accreting planet heats the disc significantly outside its Hill radius. The vortices can then migrate with respect to the planet, resulting in a complex interplay between planetary growth, gap-opening, dust trapping and vortex dynamics. Planets formed within dust traps can have masses that exceed the classical pebble isolation mass, potentially providing massive seeds for the future formation of giant planets. Once pebble accretion ceases, the local dust size distribution is depleted in large grains, and much of the remaining dust mass is trapped in the system's L$_5$ Lagrange point, providing potentially observable signatures of this evolution., Comment: 17 pages, 11 figures. Accepted for publication in MNRAS

There exists three main areas of study inside of the field of predictive maintenance: anomaly detection, fault diagnosis, and remaining useful life prediction. Notably, anomaly detection alerts the stakeholder that an anomaly is occurring. This raises two fundamental questions: what is causing the fault and how can we fix it? Inside of the field of explainable artificial intelligence, counterfactual explanations can give that information in the form of what changes to make to put the data point into the opposing class, in this case "healthy". The suggestions are not always actionable which may raise the interest in asking "what if we do this instead?" In this work, we provide a proof of concept for utilizing counterfactual explanations as what-if analysis. We perform this on the PRONOSTIA dataset with a temporal convolutional network as the anomaly detector. Our method presents the counterfactuals in the form of a what-if analysis for this base problem to inspire future work for more complex systems and scenarios., Comment: 8 pages, 6 figures, 3 tables

Sample Return Capsules (SRCs) entering Earth's atmosphere at hypervelocity from interplanetary space are a valuable resource for studying meteor phenomena. The 24 September 2023 arrival of the OSIRIS-REx (Origins, Spectral Interpretation, Resource Identification, and Security-Regolith Explorer) SRC provided an unprecedented chance for geophysical observations of a well-characterized source with known parameters, including timing and trajectory. A collaborative effort involving researchers from 16 institutions executed a carefully planned geophysical observational campaign at strategically chosen locations, deploying over 400 ground-based sensors encompassing infrasound, seismic, distributed acoustic sensing (DAS), and GPS technologies. Additionally, balloons equipped with infrasound sensors were launched to capture signals at higher altitudes. This campaign (the largest of its kind so far) yielded a wealth of invaluable data anticipated to fuel scientific inquiry for years to come. The success of the observational campaign is evidenced by the near-universal detection of signals across instruments, both proximal and distal. This paper presents a comprehensive overview of the collective scientific effort, field deployment, and preliminary findings. The early findings have the potential to inform future space missions and terrestrial campaigns, contributing to our understanding of meteoroid interactions with planetary atmospheres. Furthermore, the dataset collected during this campaign will improve entry and propagation models as well as augment the study of atmospheric dynamics and shock phenomena generated by meteoroids and similar sources., Comment: 87 pages, 14 figures

Large Language Models (LLMs) have demonstrated remarkable capabilities across a variety of software engineering and coding tasks. However, their application in the domain of code and compiler optimization remains underexplored. Training LLMs is resource-intensive, requiring substantial GPU hours and extensive data collection, which can be prohibitive. To address this gap, we introduce Meta Large Language Model Compiler (LLM Compiler), a suite of robust, openly available, pre-trained models specifically designed for code optimization tasks. Built on the foundation of Code Llama, LLM Compiler enhances the understanding of compiler intermediate representations (IRs), assembly language, and optimization techniques. The model has been trained on a vast corpus of 546 billion tokens of LLVM-IR and assembly code and has undergone instruction fine-tuning to interpret compiler behavior. LLM Compiler is released under a bespoke commercial license to allow wide reuse and is available in two sizes: 7 billion and 13 billion parameters. We also present fine-tuned versions of the model, demonstrating its enhanced capabilities in optimizing code size and disassembling from x86_64 and ARM assembly back into LLVM-IR. These achieve 77% of the optimising potential of an autotuning search, and 45% disassembly round trip (14% exact match). This release aims to provide a scalable, cost-effective foundation for further research and development in compiler optimization by both academic researchers and industry practitioners.

Synchronized behavior among individuals is a ubiquitous feature of populations. Understanding mechanisms of (de)synchronization demands meaningful, interpretable, computable quantifications of synchrony, relevant to measurements that can be made of dynamic populations. Despite the importance to analyzing and modeling populations, existing notions of synchrony often lack rigorous definitions, may be specialized to a particular experimental system and/or measurement, or may have undesirable properties that limit their utility. We introduce a notion of synchrony for populations of individuals occupying a compact metric space that depends on the Fr\'{e}chet variance of the distribution of individuals. We establish several fundamental and desirable mathematical properties of this synchrony measure, including continuity and invariance to metric scaling. We establish a general approximation result that controls the disparity between synchrony in the true space and the synchrony observed through a discretization of state space, as may occur when observable states are limited by measurement constraints. We develop efficient algorithms to compute synchrony in a variety of state spaces, including all finite state spaces and empirical distributions on the circle, and provide accessible implementations in an open-source Python module. To demonstrate the usefulness of the synchrony measure in biological applications, we investigate several biologically relevant models of mechanisms that can alter the dynamics of synchrony over time, and reanalyze published data concerning the dynamics of the intraerythrocytic developmental cycles of $\textit{Plasmodium}$ parasites. We anticipate that the rigorous definition of population synchrony and the mathematical and biological results presented here will be broadly useful in analyzing and modeling populations in a variety of contexts.

Speech and language biomarkers have the potential to be regular, objective assessments of symptom severity in several health conditions, both in-clinic and remotely using mobile devices. However, the complex nature of speech and often subtle changes associated with health mean that findings are highly dependent on methodological and cohort choices. These are often not reported adequately in studies investigating speech-based health assessment, hindering the progress of methodological speech research. Our objectives were to) facilitate replicable speech research by presenting an adaptable speech collection and analytical method and design checklist for other researchers to adapt for their own experiments and develop an exemplar protocol that reduces and controls for confounding factors in repeated recordings of speech, including device choice, speech elicitation task and non-pathological variability. The presented protocol comprises the elicitation of read speech, held vowels and a picture description collected with a freestanding condenser microphone, 3 smartphones and a headset. We extracted a set of 14 exemplar speech features. We collected healthy speech from 28 individuals 3 times in 1 day, repeated at the same times 8-11 weeks later, and from 25 individuals on 3 days in 1 week at fixed times. Participant characteristics collected included sex, age, native language status and voice use habits. Before each recording, we collected information on recent voice use, food and drink intake, and emotional state. The extracted features are presented providing a resource of normative values. Speech data collection, processing, analysis and reporting towards clinical research and practice varies widely. Greater harmonisation of study protocols and consistent reporting are urgently required to translate speech processing into clinical research and practice., Comment: Main manuscript: 37 pages. 3 figures, 8 tables, 1 textbox. Submitted to JMIR Research Methods. Replacement with format changes and copyediting

Generative AI has received much attention in the image and language domains, with the transformer neural network continuing to dominate the state of the art. Application of these models to time series generation is less explored, however, and is of great utility to machine learning, privacy preservation, and explainability research. The present survey identifies this gap at the intersection of the transformer, generative AI, and time series data, and reviews works in this sparsely populated subdomain. The reviewed works show great variety in approach, and have not yet converged on a conclusive answer to the problems the domain poses. GANs, diffusion models, state space models, and autoencoders were all encountered alongside or surrounding the transformers which originally motivated the survey. While too open a domain to offer conclusive insights, the works surveyed are quite suggestive, and several recommendations for best practice, and suggestions of valuable future work, are provided.

We explore the potential of spatial-photonic Ising machines (SPIMs) to address computationally intensive Ising problems that employ low-rank and circulant coupling matrices. Our results indicate that the performance of SPIMs is critically affected by the rank and precision of the coupling matrices. By developing and assessing advanced decomposition techniques, we expand the range of problems SPIMs can solve, overcoming the limitations of traditional Mattis-type matrices. Our approach accommodates a diverse array of coupling matrices, including those with inherently low ranks, applicable to complex NP-complete problems. We explore the practical benefits of low-rank approximation in optimization tasks, particularly in financial optimization, to demonstrate the real-world applications of SPIMs. Finally, we evaluate the computational limitations imposed by SPIM hardware precision and suggest strategies to optimize the performance of these systems within these constraints., Comment: 15 pages, 7 figures

Fitch Cheney's 5-card trick was introduced in 1950. In 2013, Mulcahy invented a 4-card trick in which the cards are allowed to be displayed face down. We suggest our own invention: a 3-card trick in which the cards can be face down and also allowed to be placed both vertically and horizontally. We discuss the theory behind all the tricks and estimate the maximum deck size given the number of chosen cards. We also discuss the cases of hiding several cards and the deck that has duplicates., Comment: 24 pages, 13 tables

Prior research has shown that changes in seasons and weather can have a significant impact on depression severity. However, findings are inconsistent across populations, and the interplay between weather, behavior, and depression has not been fully quantified. This study analyzed real-world data from 428 participants (a subset; 68.7% of the cohort) in the RADAR-MDD longitudinal mobile health study to investigate seasonal variations in depression (measured through a remote validated assessment - PHQ-8) and examine the potential interplay between dynamic weather changes, physical activity (monitored via wearables), and depression severity. The clustering of PHQ-8 scores identified four distinct seasonal variations in depression severity: one stable trend and three varying patterns where depression peaks in different seasons. Among these patterns, participants within the stable trend had the oldest average age (p=0.002) and the lowest baseline PHQ-8 score (p=0.003). Mediation analysis assessing the indirect effect of weather on physical activity and depression showed significant differences among participants with different affective responses to weather. These findings illustrate the heterogeneity in individuals' seasonal depression variations and responses to weather, underscoring the necessity for personalized approaches to help understand the impact of environmental factors on the real-world effectiveness of behavioral treatments.

Data augmentation is an important facilitator of deep learning applications in the time series domain. A gap is identified in the literature, demonstrating sparse exploration of the transformer, the dominant sequence model, for data augmentation in time series. A architecture hybridizing several successful priors is put forth and tested using a powerful time domain similarity metric. Results suggest the challenge of this domain, and several valuable directions for future work.

We introduce the Vector Ising Spin Annealer (VISA), a framework in gain-based computing that harnesses light-matter interactions to solve complex optimization problems encoded in spin Hamiltonians. Traditional driven-dissipative systems often select excited states due to limitations in spin movement. VISA transcends these constraints by enabling spins to operate in a three-dimensional space, offering a robust solution to minimize Ising Hamiltonians effectively. Our comparative analysis reveals VISA's superior performance over conventional single-dimension spin optimizers, demonstrating its ability to bridge substantial energy barriers in complex landscapes. Through detailed studies on cyclic and random graphs, we show VISA's proficiency in dynamically evolving the energy landscape with time-dependent gain and penalty annealing, illustrating its potential to redefine optimization in physical systems., Comment: 12 pages, 13 figures

We introduce a novel paradigm in compiler optimization powered by Large Language Models with compiler feedback to optimize the code size of LLVM assembly. The model takes unoptimized LLVM IR as input and produces optimized IR, the best optimization passes, and instruction counts of both unoptimized and optimized IRs. Then we compile the input with generated optimization passes and evaluate if the predicted instruction count is correct, generated IR is compilable, and corresponds to compiled code. We provide this feedback back to LLM and give it another chance to optimize code. This approach adds an extra 0.53% improvement over -Oz to the original model. Even though, adding more information with feedback seems intuitive, simple sampling techniques achieve much higher performance given 10 or more samples.

Large language models show great potential in generating and optimizing code. Widely used sampling methods such as Nucleus Sampling increase the diversity of generation but often produce repeated samples for low temperatures and incoherent samples for high temperatures. Furthermore, the temperature coefficient has to be tuned for each task, limiting its usability. We present Priority Sampling, a simple and deterministic sampling technique that produces unique samples ordered by the model's confidence. Each new sample expands the unexpanded token with the highest probability in the augmented search tree. Additionally, Priority Sampling supports generation based on regular expression that provides a controllable and structured exploration process. Priority Sampling outperforms Nucleus Sampling for any number of samples, boosting the performance of the original model from 2.87% to 5% improvement over -Oz. Moreover, it outperforms the autotuner used for the generation of labels for the training of the original model in just 30 samples.

Register allocation is one of the most important problems for modern compilers. With a practically unlimited number of user variables and a small number of CPU registers, assigning variables to registers without conflicts is a complex task. This work demonstrates the use of casting the register allocation problem as a graph coloring problem. Using technologies such as PyTorch and OpenAI Gymnasium Environments we will show that a Proximal Policy Optimization model can learn to solve the graph coloring problem. We will also show that the labeling of a graph is critical to the performance of the model by taking the matrix representation of a graph and permuting it. We then test the model's effectiveness on each of these permutations and show that it is not effective when given a relabeling of the same graph. Our main contribution lies in showing the need for label reordering invariant representations of graphs for machine learning models to achieve consistent performance.

Objective: This study aimed to explore the associations between depression severity and wearable-measured circadian rhythms, accounting for seasonal impacts and quantifying seasonal changes in circadian rhythms.Materials and Methods: Data used in this study came from a large longitudinal mobile health study. Depression severity (measured biweekly using the 8-item Patient Health Questionnaire [PHQ-8]) and behaviors (monitored by Fitbit) were tracked for up to two years. Twelve features were extracted from Fitbit recordings to approximate circadian rhythms. Three nested linear mixed-effects models were employed for each feature: (1) incorporating the PHQ-8 score as an independent variable; (2) adding the season variable; and (3) adding an interaction term between season and the PHQ-8 score. Results: This study analyzed 10,018 PHQ-8 records with Fitbit data from 543 participants. Upon adjusting for seasonal effects, higher PHQ-8 scores were associated with reduced activity, irregular behaviors, and delayed rhythms. Notably, the negative association with daily step counts was stronger in summer and spring than in winter, and the positive association with the onset of the most active continuous 10-hour period was significant only during summer. Furthermore, participants had shorter and later sleep, more activity, and delayed circadian rhythms in summer compared to winter. Discussion and Conclusions: Our findings underscore the significant seasonal impacts on human circadian rhythms and their associations with depression and indicate that wearable-measured circadian rhythms have the potential to be the digital biomarkers of depression.