The widespread adoption of Electric Vehicles (EVs) is limited by their reliance on batteries with presently low energy and power densities compared to liquid fuels and are subject to aging and performance deterioration over time. For this reason, monitoring the battery State Of Charge (SOC) and State Of Health (SOH) during the EV lifetime is a very relevant problem. This work proposes a battery digital twin structure designed to accurately reflect battery dynamics at the run time. To ensure a high degree of correctness concerning non-linear phenomena, the digital twin relies on data-driven models trained on traces of battery evolution over time: a SOH model, repeatedly executed to estimate the degradation of maximum battery capacity, and a SOC model, retrained periodically to reflect the impact of aging. The proposed digital twin structure will be exemplified on a public dataset to motivate its adoption and prove its effectiveness, with high accuracy and inference and retraining times compatible with onboard execution.
We experimentally validate a machine learning-enabled Raman amplification framework, capable of jointly shaping the signal power evolution in two domains: frequency and fiber distance. The proposed experiment addresses the amplification in the whole C-band, by optimizing four first-order counter-propagating Raman pumps.
Hyperparameter optimization constitutes a large part of typical modern machine learning workflows. This arises from the fact that machine learning methods and corresponding preprocessing steps often only yield optimal performance when hyperparameters are properly tuned. But in many applications, we are not only interested in optimizing ML pipelines solely for predictive accuracy; additional metrics or constraints must be considered when determining an optimal configuration, resulting in a multi-objective optimization problem. This is often neglected in practice, due to a lack of knowledge and readily available software implementations for multi-objective hyperparameter optimization. In this work, we introduce the reader to the basics of multi-objective hyperparameter optimization and motivate its usefulness in applied ML. Furthermore, we provide an extensive survey of existing optimization strategies, both from the domain of evolutionary algorithms and Bayesian optimization. We illustrate the utility of MOO in several specific ML applications, considering objectives such as operating conditions, prediction time, sparseness, fairness, interpretability and robustness.
Nonlinear wave interactions, such as shock refraction at an inclined density interface, in magnetohydrodynamic (MHD) lead to a plethora of wave patterns with myriad wave types. Identification of different types of MHD waves is an important and challenging task in such complex wave patterns. Moreover, owing to the multiplicity of solutions and their admissibility for different systems, especially for intermediate-type MHD shock waves, the identification of MHD wave types is complicated if one solely relies on the Rankine-Hugoniot jump conditions. MHD wave detection is further exacerbated by the unphysical smearing of discontinuous shock waves in numerical simulations. We present two MHD wave detection methods based on a convolutional neural network (CNN) which enables the classification of waves and identification of their locations. The first method separates the output into a regression (location prediction) and a classification problem assuming the number of waves for each training data is fixed. In the second method, the number of waves is not specified a priori and the algorithm, using only regression, predicts the waves' locations and classifies their types. The first fixed output model efficiently provides high precision and recall, the accuracy of the entire neural network achieved is up to 0.99, and the classification accuracy of some waves approaches unity. The second detection model has relatively lower performance, with more sensitivity to the setting of parameters, such as the number of grid cells N_{grid} and the thresholds of confidence score and class probability, etc. The proposed two methods demonstrate very strong potential to be applied for MHD wave detection in some complex wave structures and interactions.
Modern applications such as voice recognition rely on the ability to compare signals to pre-recorded ones to classify them. However, this comparison typically needs to ignore differences due to signal noise, temporal offset, signal magnitude, and other external factors. The Dynamic Time Warping (DTW) algorithm quantifies this similarity by finding corresponding regions between the signals and non-linearly warping one signal by stretching and shrinking it. Unfortunately, searching through all "warps" of a signal to find the best corresponding regions is computationally expensive. The FastDTW algorithm improves performance, but sacrifices accuracy by only considering small signal warps. My goal is to improve the speed of DTW while maintaining high accuracy. My key insight is that in any particular application domain, signals exhibit specific types of variation. For example, the accelerometer signal measured for two different people would differ based on their stride length and weight. My system, called Machine Learning DTW (MLDTW), uses machine learning to learn the types of warps that are common in a particular domain. It then uses the learned model to improve DTW performance by limiting the search of potential warps appropriately. My results show that compared to FastDTW, MLDTW is at least as fast and reduces errors by 60% on average across four different data sets. These improvements will significantly impact a wide variety of applications (e.g. health monitoring) and enable more scalable processing of multivariate, higher frequency, and longer signal recordings.
Knowledge of mixtures' phase equilibria is crucial in nature and technical chemistry. Phase equilibria calculations of mixtures require activity coefficients. However, experimental data on activity coefficients is often limited due to high cost of experiments. For an accurate and efficient prediction of activity coefficients, machine learning approaches have been recently developed. However, current machine learning approaches still extrapolate poorly for activity coefficients of unknown molecules. In this work, we introduce the SMILES-to-Properties-Transformer (SPT), a natural language processing network to predict binary limiting activity coefficients from SMILES codes. To overcome the limitations of available experimental data, we initially train our network on a large dataset of synthetic data sampled from COSMO-RS (10 Million data points) and then fine-tune the model on experimental data (20 870 data points). This training strategy enables SPT to accurately predict limiting activity coefficients even for unknown molecules, cutting the mean prediction error in half compared to state-of-the-art models for activity coefficient predictions such as COSMO-RS, UNIFAC, and improving on recent machine learning approaches.
Application of machine learning may be understood as deriving new knowledge for practical use through explaining accumulated observations, training set. Peirce used the term abduction for this kind of inference. Here I formalize the concept of abduction for real valued hypotheses, and show that 14 of the most popular textbook ML learners (every learner I tested), covering classification, regression and clustering, implement this concept of abduction inference. The approach is proposed as an alternative to statistical learning theory, which requires an impractical assumption of indefinitely increasing training set for its justification.
Controlling infectious diseases is a major health priority because they can spread and infect humans, thus evolving into epidemics or pandemics. Therefore, early detection of infectious diseases is a significant need, and many researchers have developed models to diagnose them in the early stages. This paper reviewed research articles for recent machine-learning (ML) algorithms applied to infectious disease diagnosis. We searched the Web of Science, ScienceDirect, PubMed, Springer, and IEEE databases from 2015 to 2022, identified the pros and cons of the reviewed ML models, and discussed the possible recommendations to advance the studies in this field. We found that most of the articles used small datasets, and few of them used real-time data. Our results demonstrated that a suitable ML technique depends on the nature of the dataset and the desired goal.
Fine-tuning pretrained language models (LMs) without making any architectural changes has become a norm for learning various language downstream tasks. However, for non-language downstream tasks, a common practice is to employ task-specific designs for input, output layers, and loss functions. For instance, it is possible to fine-tune an LM into an MNIST classifier by replacing the word embedding layer with an image patch embedding layer, the word token output layer with a 10-way output layer, and the word prediction loss with a 10-way classification loss, respectively. A natural question arises: Can LM fine-tuning solve non-language downstream tasks without changing the model architecture or loss function? To answer this, we propose Language-Interfaced Fine-Tuning (LIFT) and study its efficacy and limitations by conducting an extensive empirical study on a suite of non-language classification and regression tasks. LIFT does not make any changes to the model architecture or loss function, and it solely relies on the natural language interface, enabling "no-code machine learning with LMs." We find that LIFT performs comparably well across a wide range of low-dimensional classification and regression tasks, matching the performances of the best baselines in many cases, especially for the classification tasks. We also report experimental results on the fundamental properties of LIFT, including inductive bias, robustness, and sample complexity. We also analyze the effect of pretraining on LIFT and a few properties/techniques specific to LIFT, e.g., context-aware learning via appropriate prompting, calibrated predictions, data generation, and two-stage fine-tuning. Our code is available at https://github.com/UW-Madison-Lee-Lab/LanguageInterfacedFineTuning.
Advances in neuroimaging techniques have provided us novel insights into understanding how the human mind works. Functional magnetic resonance imaging (fMRI) is the most popular and widely used neuroimaging technique, and there is growing interest in fMRI-based markers of individual differences. However, its utility is often limited due to its high cost and difficulty acquiring from specific populations, including children and infants. Surrogate markers, or neural correlates of fMRI markers, would have important practical implications, but we have few stand-alone predictors for the fMRI markers. Here, using machine learning (ML) models and data augmentation, we predicted well-validated fMRI markers of human cognition from multivariate patterns of functional near-infrared spectroscopy (fNIRS), a portable and relatively inexpensive optical neuroimaging technique. We recruited 50 human participants who performed two cognitive tasks (stop signal task and probabilistic reversal learning task), while neural activation was measured with either fNIRS or fMRI at each of the total two visits. Using ML models and data augmentation, we could predict the well-established fMRI markers of response inhibition or prediction error signals from 48-channel fNIRS activation in the prefrontal cortex. These results suggest that fNIRS might offer a surrogate marker of fMRI activation, which would broaden our understanding of various populations, including infants.
The anonymous nature of darknets is commonly exploited for illegal activities. Previous research has employed machine learning and deep learning techniques to automate the detection of darknet traffic in an attempt to block these criminal activities. This research aims to improve darknet traffic detection by assessing Support Vector Machines (SVM), Random Forest (RF), Convolutional Neural Networks (CNN), and Auxiliary-Classifier Generative Adversarial Networks (AC-GAN) for classification of such traffic and the underlying application types. We find that our RF model outperforms the state-of-the-art machine learning techniques used in prior work with the CIC-Darknet2020 dataset. To evaluate the robustness of our RF classifier, we obfuscate select application type classes to simulate realistic adversarial attack scenarios. We demonstrate that our best-performing classifier can be defeated by such attacks, and we consider ways to deal with such adversarial attacks.
The idealization of a static machine-learned model, trained once and deployed forever, is not practical. As input distributions change over time, the model will not only lose accuracy, any constraints to reduce bias against a protected class may fail to work as intended. Thus, researchers have begun to explore ways to maintain algorithmic fairness over time. One line of work focuses on dynamic learning: retraining after each batch, and the other on robust learning which tries to make algorithms robust against all possible future changes. Dynamic learning seeks to reduce biases soon after they have occurred and robust learning often yields (overly) conservative models. We propose an anticipatory dynamic learning approach for correcting the algorithm to mitigate bias before it occurs. Specifically, we make use of anticipations regarding the relative distributions of population subgroups (e.g., relative ratios of male and female applicants) in the next cycle to identify the right parameters for an importance weighing fairness approach. Results from experiments over multiple real-world datasets suggest that this approach has promise for anticipatory bias correction.
Branch-and-bound is a systematic enumerative method for combinatorial optimization, where the performance highly relies on the variable selection strategy. State-of-the-art handcrafted heuristic strategies suffer from relatively slow inference time for each selection, while the current machine learning methods require a significant amount of labeled data. We propose a new approach for solving the data labeling and inference latency issues in combinatorial optimization based on the use of the reinforcement learning (RL) paradigm. We use imitation learning to bootstrap an RL agent and then use Proximal Policy Optimization (PPO) to further explore global optimal actions. Then, a value network is used to run Monte-Carlo tree search (MCTS) to enhance the policy network. We evaluate the performance of our method on four different categories of combinatorial optimization problems and show that our approach performs strongly compared to the state-of-the-art machine learning and heuristics based methods.
Predictive machine learning models nowadays are often updated in a stateless and expensive way. The two main future trends for companies that want to build machine learning-based applications and systems are real-time inference and continual updating. Unfortunately, both trends require a mature infrastructure that is hard and costly to realize on-premise. This paper defines a novel software service and model delivery infrastructure termed Continual Learning-as-a-Service (CLaaS) to address these issues. Specifically, it embraces continual machine learning and continuous integration techniques. It provides support for model updating and validation tools for data scientists without an on-premise solution and in an efficient, stateful and easy-to-use manner. Finally, this CL model service is easy to encapsulate in any machine learning infrastructure or cloud system. This paper presents the design and implementation of a CLaaS instantiation, called LiquidBrain, evaluated in two real-world scenarios. The former is a robotic object recognition setting using the CORe50 dataset while the latter is a named category and attribute prediction using the DeepFashion-C dataset in the fashion domain. Our preliminary results suggest the usability and efficiency of the Continual Learning model services and the effectiveness of the solution in addressing real-world use-cases regardless of where the computation happens in the continuum Edge-Cloud.
The deformation of cellular membranes regulates trafficking processes, such as exocytosis and endocytosis. Classically, the Helfrich continuum model is used to characterize the forces and mechanical parameters that cells tune to accomplish membrane shape changes. While this classical model effectively captures curvature generation, one of the core challenges in using it to approximate a biological process is selecting a set of mechanical parameters (including bending modulus and membrane tension) from a large set of reasonable values. We used the Helfrich model to generate a large synthetic dataset from a random sampling of realistic mechanical parameters and used this dataset to train machine learning models. These models produced promising results, accurately classifying model behavior and predicting membrane shape from mechanical parameters. We also note emerging methods in machine learning that can leverage the physical insight of the Helfrich model to improve performance and draw greater insight into how cells control membrane shape change.
Machine learning (ML) is often viewed as a black-box regression technique that is unable to provide considerable scientific insight. ML models are universal function approximators and - if used correctly - can provide scientific information related to the ground-truth dataset used for fitting. A benefit to ML over parametric models is that there are no predefined basis functions limiting the phenomena that can be modeled. In this work, we develop ML models on three datasets: the Space Environment Technologies (SET) High Accuracy Satellite Drag Model (HASDM) density database, a spatiotemporally matched dataset of outputs from the Jacchia-Bowman 2008 Empirical Thermospheric Density Model (JB2008), and an accelerometer-derived density dataset from CHAllenging Minisatellite Payload (CHAMP). These ML models are compared to the Naval Research Laboratory Mass Spectrometer and Incoherent Scatter radar (NRLMSIS 2.0) model to study the presence of post-storm cooling in the middle-thermosphere. We find that both NRLMSIS 2.0 and JB2008-ML do not account for post-storm cooling and consequently perform poorly in periods following strong geomagnetic storms (e.g. the 2003 Halloween storms). Conversely, HASDM-ML and CHAMP-ML do show evidence of post-storm cooling indicating that this phenomenon is present in the original datasets. Results show that density reductions up to 40% can occur 1--3 days post-storm depending on location and the strength of the storm.
We consider the use of machine learning for hypothesis testing with an emphasis on target detection. Classical model-based solutions rely on comparing likelihoods. These are sensitive to imperfect models and are often computationally expensive. In contrast, data-driven machine learning is often more robust and yields classifiers with fixed computational complexity. Learned detectors usually provide high accuracy with low complexity but do not have a constant false alarm rate (CFAR) as required in many applications. To close this gap, we propose to add a term to the loss function that promotes similar distributions of the detector under any null hypothesis scenario. Experiments show that our approach leads to near CFAR detectors with similar accuracy as their competitors.
The computationally expensive estimation of engineering demand parameters (EDPs) via finite element (FE) models, while considering earthquake and parameter uncertainty limits the use of the Performance Based Earthquake Engineering framework. Attempts have been made to substitute FE models with surrogate models, however, most of these models are a function of building parameters only. This necessitates re-training for earthquakes not previously seen by the surrogate. In this paper, the authors propose a machine learning based surrogate model framework, which considers both these uncertainties in order to predict for unseen earthquakes. Accordingly,earthquakes are characterized by their projections on an orthonormal basis, computed using SVD of a representative ground motion suite. This enables one to generate large varieties of earthquakes by randomly sampling these weights and multiplying them with the basis. The weights along with the constitutive parameters serve as inputs to a machine learning model with EDPs as the desired output. Four competing machine learning models were tested and it was observed that a deep neural network (DNN) gave the most accurate prediction. The framework is validated by using it to successfully predict the peak response of one-story and three-story buildings represented using stick models, subjected to unseen far-field ground motions.
To learn about real world phenomena, scientists have traditionally used models with clearly interpretable elements. However, modern machine learning (ML) models, while powerful predictors, lack this direct elementwise interpretability (e.g. neural network weights). Interpretable machine learning (IML) offers a solution by analyzing models holistically to derive interpretations. Yet, current IML research is focused on auditing ML models rather than leveraging them for scientific inference. Our work bridges this gap, presenting a framework for designing IML methods-termed 'property descriptors' -- that illuminate not just the model, but also the phenomenon it represents. We demonstrate that property descriptors, grounded in statistical learning theory, can effectively reveal relevant properties of the joint probability distribution of the observational data. We identify existing IML methods suited for scientific inference and provide a guide for developing new descriptors with quantified epistemic uncertainty. Our framework empowers scientists to harness ML models for inference, and provides directions for future IML research to support scientific understanding.
A final exam in machine learning at a top institution such as MIT, Harvard, or Cornell typically takes faculty days to write, and students hours to solve. We demonstrate that large language models pass machine learning finals at a human level, on finals available online after the models were trained, and automatically generate new human-quality final exam questions in seconds. Previous work has developed program synthesis and few-shot learning methods to solve university-level problem set questions in mathematics and STEM courses. In this work, we develop and compare methods that solve final exams, which differ from problem sets in several ways: the questions are longer, have multiple parts, are more complicated, and span a broader set of topics. We curate a dataset and benchmark of questions from machine learning final exams available online and code for answering these questions and generating new questions. We show how to generate new questions from other questions and course notes. For reproducibility and future research on this final exam benchmark, we use automatic checkers for multiple-choice, numeric, and questions with expression answers. We perform ablation studies comparing zero-shot learning with few-shot learning and chain-of-thought prompting using GPT-3, OPT, Codex, and ChatGPT across machine learning topics and find that few-shot learning methods perform best. We highlight the transformative potential of language models to streamline the writing and solution of large-scale assessments, significantly reducing the workload from human days to mere machine seconds. Our results suggest that rather than banning large language models such as ChatGPT in class, instructors should teach students to harness them by asking students meta-questions about correctness, completeness, and originality of the responses generated, encouraging critical thinking in academic studies.