Impossibility theorems for feature attribution

Despite a sea of interpretability methods that can produce plausible explanations, the field has also empirically seen many failure cases of such methods. In light of these results, it remains unclear for practitioners how to use these methods and choose between them in a principled way. In this paper, we show that for moderately rich model classes (easily satisfied by neural networks), any feature attribution method that is complete and linear-for example, Integrated Gradients and Shapley Additive Explanations (SHAP)-can provably fail to improve on random guessing for inferring model behavior. Our results apply to common end-tasks such as characterizing local model behavior, identifying spurious features, and algorithmic recourse. One takeaway from our work is the importance of concretely defining end-tasks: Once such an end-task is defined, a simple and direct approach of repeated model evaluations can outperform many other complex feature attribution methods.

The Effect of Lameness on Milk Production of Dairy Goats

Lameness on dairy goat farms is a welfare concern and could negatively affect milk production. This study's objective was to evaluate the effects of clinical lameness on the daily milk production of dairy goats. Between July 2019 and June 2020, 11,847 test-day records were collected from 3145 goats on three farms in New Zealand. Locomotion scoring of goats used a five-point scoring system (0 to 4). The dataset was split into two groups by lactation type, where goats were classified as being in seasonal lactation (≤305 days in milk) or extended lactation (>305 days in milk). A linear mixed model was used to analyze datasets using milk characteristics as the dependent variables. Severely lame goats (score 4) in seasonal and extended lactation produced 7.05% and 8.67% less milk than goats not lame, respectively. When the prevalence of severe lameness is between 5 and 20% of the herd, the estimated average daily milk income lost was between NZD 19.5 and 104 per day. This study established the negative impact of lameness on milk production and annual income in dairy goats on three farms.

Estimates of Genetic Parameters for Milk, the Occurrence of and Susceptibility to Clinical Lameness and Claw Disorders in Dairy Goats

The New Zealand goat industry accesses niche markets for high-value products, mainly formula for infants and young children. This study aimed to estimate the genetic parameters of occurrence and susceptibility of clinical lameness and selected claw disorders and establish their genetic associations with milk production traits. Information on pedigree, lameness, claw disorders, and milk production was collected on three farms between June 2019 and July 2020. The dataset contained 1637 does from 174 sires and 1231 dams. Estimates of genetic and residual (co)variances, heritabilities, and genetic and phenotypic correlations were obtained with uni- and bi-variate animal models. The models included the fixed effects of farm and parity, deviation from the median kidding date as a covariate, and the random effects of animal and residual error. The heritability (h²) estimates for lameness occurrence and susceptibility were 0.07 and 0.13, respectively. The h² estimates for claw disorder susceptibilities ranged from 0.02 to 0.23. The genotypic correlations ranged from weak to very strong between lameness and milk production traits (-0.94 to 0.84) and weak to moderate (0.23 to 0.84) between claw disorder and milk production traits.

Wearables: An R Package With Accompanying Shiny Application for Signal Analysis of a Wearable Device Targeted at Clinicians and Researchers

Physiological signals (e.g., heart rate, skin conductance) that were traditionally studied in neuroscientific laboratory research are currently being used in numerous real-life studies using wearable technology. Physiological signals obtained with wearables seem to offer great potential for continuous monitoring and providing biofeedback in clinical practice and healthcare research. The physiological data obtained from these signals has utility for both clinicians and researchers. Clinicians are typically interested in the day-to-day and moment-to-moment physiological reactivity of patients to real-life stressors, events, and situations or interested in the physiological reactivity to stimuli in therapy. Researchers typically apply signal analysis methods to the data by pre-processing the physiological signals, detecting artifacts, and extracting features, which can be a challenge considering the amount of data that needs to be processed. This paper describes the creation of a “Wearables” R package and a Shiny “E4 dashboard” application for an often-studied wearable, the Empatica E4. The package and Shiny application can be used to visualize the relationship between physiological signals and real-life stressors or stimuli, but can also be used to pre-process physiological data, detect artifacts, and extract relevant features for further analysis. In addition, the application has a batch process option to analyze large amounts of physiological data into ready-to-use data files. The software accommodates users with a downloadable report that provides opportunities for a careful investigation of physiological reactions in daily life. The application is freely available, thought to be easy to use, and thought to be easily extendible to other wearable devices. Future research should focus on the usability of the application and the validation of the algorithms.

Personalized Multitask Learning for Predicting Tomorrow's Mood, Stress, and Health

While accurately predicting mood and wellbeing could have a number of important clinical benefits, traditional machine learning (ML) methods frequently yield low performance in this domain. We posit that this is because a one-size-fits-all machine learning model is inherently ill-suited to predicting outcomes like mood and stress, which vary greatly due to individual differences. Therefore, we employ Multitask Learning (MTL) techniques to train personalized ML models which are customized to the needs of each individual, but still leverage data from across the population. Three formulations of MTL are compared: i) MTL deep neural networks, which share several hidden layers but have final layers unique to each task; ii) Multi-task Multi-Kernel learning, which feeds information across tasks through kernel weights on feature types; and iii) a Hierarchical Bayesian model in which tasks share a common Dirichlet Process prior. We offer the code for this work in open source. These techniques are investigated in the context of predicting future mood, stress, and health using data collected from surveys, wearable sensors, smartphone logs, and the weather. Empirical results demonstrate that using MTL to account for individual differences provides large performance improvements over traditional machine learning methods and provides personalized, actionable insights.

Vomit Comet Physiology: Autonomic Changes in Novice Flyers

This exploratory study examined the effects of varying g-forces, including feelings of weightlessness, on an individual's physiology during parabolic flight. Specifically, we collected heart rate, accelerometer, and skin conductance measurements from 16 flyers aboard a parabolic flight using wearable, wireless sensors. The biosignals were then correlated to participant reports of nausea, anxiety, and excitement during periods of altered g-forces. Using linear mixed-effects models, we found that (1) heart rate was positively correlated to individuals' self-reported highest/lowest periods of both anxiety and excitement, and (2) bilateral skin conductance asymmetry was positively correlated to individuals' self-reported highest/lowest periods of nausea.

Wavelet-based motion artifact removal for electrodermal activity

Electrodermal activity (EDA) recording is a powerful, widely used tool for monitoring psychological or physiological arousal. However, analysis of EDA is hampered by its sensitivity to motion artifacts. We propose a method for removing motion artifacts from EDA, measured as skin conductance (SC), using a stationary wavelet transform (SWT). We modeled the wavelet coefficients as a Gaussian mixture distribution corresponding to the underlying skin conductance level (SCL) and skin conductance responses (SCRs). The goodness-of-fit of the model was validated on ambulatory SC data. We evaluated the proposed method in comparison with three previous approaches. Our method achieved a greater reduction of artifacts while retaining motion-artifact-free data.

Automatic identification of artifacts in electrodermal activity data

Recently, wearable devices have allowed for long term, ambulatory measurement of electrodermal activity (EDA). Despite the fact that ambulatory recording can be noisy, and recording artifacts can easily be mistaken for a physiological response during analysis, to date there is no automatic method for detecting artifacts. This paper describes the development of a machine learning algorithm for automatically detecting EDA artifacts, and provides an empirical evaluation of classification performance. We have encoded our results into a freely available web-based tool for artifact and peak detection.

Predicting students' happiness from physiology, phone, mobility, and behavioral data

In order to model students' happiness, we apply machine learning methods to data collected from undergrad students monitored over the course of one month each. The data collected include physiological signals, location, smartphone logs, and survey responses to behavioral questions. Each day, participants reported their wellbeing on measures including stress, health, and happiness. Because of the relationship between happiness and depression, modeling happiness may help us to detect individuals who are at risk of depression and guide interventions to help them. We are also interested in how behavioral factors (such as sleep and social activity) affect happiness positively and negatively. A variety of machine learning and feature selection techniques are compared, including Gaussian Mixture Models and ensemble classification. We achieve 70% classification accuracy of self-reported happiness on held-out test data.

Recognizing Academic Performance, Sleep Quality, Stress Level, and Mental Health using Personality Traits, Wearable Sensors and Mobile Phones

What can wearable sensors and usage of smart phones tell us about academic performance, self-reported sleep quality, stress and mental health condition? To answer this question, we collected extensive subjective and objective data using mobile phones, surveys, and wearable sensors worn day and night from 66 participants, for 30 days each, totaling 1,980 days of data. We analyzed daily and monthly behavioral and physiological patterns and identified factors that affect academic performance (GPA), Pittsburg Sleep Quality Index (PSQI) score, perceived stress scale (PSS), and mental health composite score (MCS) from SF-12, using these month-long data. We also examined how accurately the collected data classified the participants into groups of high/low GPA, good/poor sleep quality, high/low self-reported stress, high/low MCS using feature selection and machine learning techniques. We found associations among PSQI, PSS, MCS, and GPA and personality types. Classification accuracies using the objective data from wearable sensors and mobile phones ranged from 67-92%.

Prediction of Happy-Sad mood from daily behaviors and previous sleep history

We collected and analyzed subjective and objective data using surveys and wearable sensors worn day and night from 68 participants for ~30 days each, to address questions related to the relationships among sleep duration, sleep irregularity, self-reported Happy-Sad mood and other daily behavioral factors in college students. We analyzed this behavioral and physiological data to (i) identify factors that classified the participants into Happy-Sad mood using support vector machines (SVMs); and (ii) analyze how accurately sleep duration and sleep regularity for the past 1-5 days classified morning Happy-Sad mood. We found statistically significant associations amongst Sad mood and poor health-related factors. Behavioral factors including the frequency of negative social interactions, and negative emails, and total academic activity hours showed the best performance in separating the Happy-Sad mood groups. Sleep regularity and sleep duration predicted daily Happy-Sad mood with 65-80% accuracy. The number of nights giving the best prediction of Happy-Sad mood varied for different individuals.