Maximizing Signal in Human-Model Preference Alignment

This paper studies the performance of classical and quantum machine learning models for sentiment analysis task. Here, popular machine learning algorithms viz support vector machine (SVM), gradient boosting (GB) and random forest (RF) are compared with variational quantum classifier (VQC) using two sets of parameterized circuits viz EfficientSU2 and RealAmplitudes. For experimenting with VQC, IBM Quantum Experience and IBM Qiskit were used while for classical machine learning models, scikit-learn was used. It was found that the performance of the VQC was slightly better than popular machine learning algorithms. For our experiments, we have used popular restaurant sentiment analysis dataset. The extracted features from this dataset and then after applying PCA reduced the feature set into 5 features. Quantum ML models were trained using 100 epochs and 150 epochs. Overall, four Quantum ML models were trained and three Classical ML models were trained. The performance of the trained models was evaluated using standard evaluation measures viz, Accuracy, Precision, Recall, F-Score etc. In all the cases EfficientSU2 based model with 100 Epochs performed better than all other models. Efficient SU2 model with 100 epochs produced an accuracy of 74.5% and an F-Score of 0.7605 which were highest across all the trained models.

In this chapter, we introduce several applications of machine learning and deep learning in different domains, including sensor and time-series, image and vision, text and natural language processing, relational data, energy, manufacturing, social media, health, security, and Internet-of-Things (IoT) applications. Until 2010, traditional ML models such as SVMs and decision trees have enjoyed successes in various tasks, including handwritten digit classification, face detection, and pattern recognition. Though traditional ML models are easy to interpret, the model’s inputs need to be well-designed, handcrafted features. On the other hand, deep learning models circumvent this problem and directly take the raw data as input and provide end-to-end learning capability. There is an unprecedented increase in machine learning and deep learning applications, especially with the emergence of fast mobile devices with access to cloud computing. While cloud computing provides the necessary computational power to train deep learning models, trained models can be easily deployed in the cloud or on embedded devices at the edge of the cloud to carry out the inference.

We present Ultron-AutoML, an open-source, distributed framework for efficient hyper-parameter optimization (HPO) of ML models. Considering that hyper-parameter optimization is compute intensive and time-consuming, the framework has been designed for reliability – the ability to successfully complete an HPO Job in a multi-tenant, failure prone environment, as well as efficiency – completing the job with minimum compute cost and wall-clock time. From a user’s perspective, the framework emphasizes ease of use and customizability. The user can declaratively specify and execute an HPO Job, while ancillary tasks – containerizing and running the user’s scripts, model checkpointing, monitoring progress, parallelization – are handled by the framework. At the same time, the user has complete flexibility in composing the code-base for specifying the ML model training algorithm as well as, optionally, any custom HPO algorithm. The framework supports the creation of data-pipelines to stream batches of shuffled and augmented data from a distributed file system. This comes in handy for training Deep Learning models based on self-supervised, semi-supervised or representation learning algorithms over large training datasets. We demonstrate the framework’s reliability and efficiency by running a BERT pre-training job over a large training corpus using pre-emptible GPU compute targets. Despite the inherent unreliability of the underlying compute nodes, the framework is able to complete such long running jobs at 30% of the cost with a marginal increase in wall-clock time. The framework also comes with a service to monitor jobs and ensures reproducibility of any result.

Anomaly detection is increasingly important to handle the amount of sensor data in Edge and Fog environments, Smart Cities, as well as in Industry 4.0. To ensure good results, the utilized ML models need to be updated periodically to adapt to seasonal changes and concept drifts in the sensor data. Although the increasing resource availability at the edge can allow for in-situ execution of model training directly on the devices, it is still often offloaded to fog devices or the cloud. In this paper, we propose Local-Optimistic Scheduling (LOS), a method for executing periodic ML model training jobs in close proximity to the data sources, without overloading lightweight edge devices. Training jobs are offloaded to nearby neighbor nodes as necessary and the resource consumption is optimized to meet the training period while still ensuring enough resources for further training executions. This scheduling is accomplished in a decentralized, collaborative and opportunistic manner, without full knowledge of the infrastructure and workload. We evaluated our method in an edge computing testbed on real-world datasets. The experimental results show that LOS places the training executions close to the input sensor streams, decreases the deviation between training time and training period by up to 40% and increases the amount of successfully scheduled training jobs compared to an in-situ execution.

Machine learning (ML) has been pervasively researched nowadays and it has been applied in many aspects of real life. Nevertheless, issues of model and data still accompany the development of ML. For instance, training of traditional ML models is limited to the access of data sets, which are generally proprietary; published ML models may soon be out of date without an update of new data and continuous training; malicious data contributors may upload wrongly labeled data that leads to undesirable training results; and the abuse of private data and data leakage also exist. With the utilization of blockchain, an emerging and swiftly developing technology, these problems can be efficiently solved. In this paper, we survey the convergence of collaborative ML and blockchain. Different ways of the combination of these two technologies are investigated and their fields of application are examined. Discussion on the limitations of current research and their future directions are also included. KeywordsMachine Learning (ML)BlockchainMarketplaceSmart contractIncentive mechanismData sharingModel sharingData privacyTransaction privacy

Ensemble methods are state of the art for many NLP tasks. Recent work by Banko and Brill (2001) suggests that this would not necessarily be true if very large training corpora were available. However, their results are limited by the simplicity of their evaluation task and individual classifiers.Our work explores ensemble efficacy for the more complex task of automatic thesaurus extraction on up to 300 million words. We examine our conflicting results in terms of the constraints on, and complexity of, different contextual representations, which contribute to the sparseness-and noise-induced bias behaviour of NLP systems on very large corpora.

Existing reference-free metrics have obvious limitations for evaluating controlled text generation models. Unsupervised metrics can only provide a task-agnostic evaluation result which correlates weakly with human judgments, whereas supervised ones may overfit task-specific data with poor generalization ability to other datasets. In this paper, we propose an unsupervised reference-free metric called CTRLEval, which evaluates controlled text generation from different aspects by formulating each aspect into multiple text infilling tasks. On top of these tasks, the metric assembles the generation probabilities from a pre-trained language model without any model training. Experimental results show that our metric has higher correlations with human judgments than other baselines, while obtaining better generalization of evaluating generated texts from different models and with different qualities.

Natural language understanding (NLU) and natural language generation (NLG) are two fundamental and related tasks in building task-oriented dialogue systems with opposite objectives: NLU tackles the transformation from natural language to formal representations, whereas NLG does the reverse. A key to success in either task is parallel training data which is expensive to obtain at a large scale. In this work, we propose a generative model which couples NLU and NLG through a shared latent variable. This approach allows us to explore both spaces of natural language and formal representations, and facilitates information sharing through the latent space to eventually benefit NLU and NLG. Our model achieves state-of-the-art performance on two dialogue datasets with both flat and tree-structured formal representations. We also show that the model can be trained in a semi-supervised fashion by utilising unlabelled data to boost its performance.

Self-supervised pre-training, such as BERT, MASS and BART, has emerged as a powerful technique for natural language understanding and generation. Existing pre-training techniques employ autoencoding and/or autoregressive objectives to train Transformer-based models by recovering original word tokens from corrupted text with some masked tokens. The training goals of existing techniques are often inconsistent with the goals of many language generation tasks, such as generative question answering and conversational response generation, for producing new text given context. This work presents PALM with a novel scheme that jointly pre-trains an autoencoding and autoregressive language model on a large unlabeled corpus, specifically designed for generating new text conditioned on context. The new scheme alleviates the mismatch introduced by the existing denoising scheme between pre-training and fine-tuning where generation is more than reconstructing original text. An extensive set of experiments show that PALM achieves new state-of-the-art results on a variety of language generation benchmarks covering generative question answering (Rank 1 on the official MARCO leaderboard), abstractive summarization on CNN/DailyMail as well as Gigaword, question generation on SQuAD, and conversational response generation on Cornell Movie Dialogues.

The defining property of a Concept-to-Speech system is that it combines language and speech generation. Language generation converts the input concepts into natural language, which speech generation subsequently transforms into speech. Potentially, this leads to a more ‘natural sounding’ output than can be achieved in a plain Text-to-Speech system, since the correct placement of pitch accents and intonational boundaries —an important factor contributing to the ‘naturalness’ of the generated speech— is co-determined by syntactic and discourse information, which is typically available in the language generation module. In this paper, a generic algorithm for the generation of coherent spoken monologues is discussed, called D2S. Language generation is done by a module called LGM which is based on TAG-like syntactic structures with open slots, combined with conditions which determine when the syntactic structure can be used properly. A speech generation module (SGM) converts the output of the LGM into speech using either phrase-concatenation or diphone-synthesis.

Machine Learning Research often involves the use of diverse libraries, modules, and pseudocodes for data processing, cleaning, filtering, pattern recognition, and computer intelligence. Quantization of Effort Required for the above cumulative processes is rarely discussed in the existing works & The time to reach the desired level of the model’s functionality is essential to gauge the environment training for the model training & pre-deployment testing. In this study, we empirically defined the manual-cum-computational effort required for the model development in terms of 2-time factors: time-to-live(time until 1st code modification) and time-for-modification(time between 2 codebase changes) taken together in a mathematical model to compute Effort Factor(quantitative measure of determining effort needed to design ML algorithms). The study is novel in terms of how the effort required to create ML models is calculated with a particular focus on the manual effort direction. The results and the findings obtained can be used for determining the total time needed to synthesize ML models & frameworks in terms of code change cycles and implementation strategies marking a 25% performance increase in the current method above the Standard Pipeline.

The use of machine learning techniques in computational chemistry has gained significant momentum since large molecular databases are now readily available. Predictions of molecular properties using machine learning have advantages over the traditional quantum mechanics calculations because they can be cheaper computationally without losing the accuracy. We present a new extrapolatable and explainable molecular representation based on bonds, angles and dihedrals that can be used to train machine learning models. The trained models can accurately predict the electronic energy and the free energy of small organic molecules with atom types C, H N and O, with a mean absolute error of 1.2 kcal mol−1. The models can be extrapolated to larger organic molecules with an average error of less than 3.7 kcal mol−1 for 10 or fewer heavy atoms, which represent a chemical space two orders of magnitude larger. The rapid energy predictions of multiple molecules, up to 7 times faster than previous ML models of similar accuracy, has been achieved by sampling geometries around the potential energy surface minima. Therefore, the input geometries do not have to be located precisely on the minima and we show that accurate density functional theory energy predictions can be made from force-field optimised geometries with a mean absolute error 2.5 kcal mol−1.

The application of machine learning techniques in seismology has greatly advanced seismological analysis, especially for earthquake detection and seismic phase picking. However, machine learning approaches still face challenges in generalizing to data sets that differ from their original training setting. Previous studies focused on retraining or transfer‐learning models for these scenarios, but require high‐quality labeled data sets. This paper demonstrates a new approach for augmenting already trained models without the need for additional training data. We propose four strategies—rescaling, model aggregation, shifting, and filtering—to enhance the performance of pre‐trained models on out‐of‐distribution data sets. We further devise various methodologies to ensemble the individual predictions from these strategies to obtain a final unified prediction result featuring prediction robustness and detection sensitivity. We develop an open‐source Python module quakephase that implements these methods and can flexibly process input continuous seismic data of any sampling rate. With quakephase and pre‐trained ML models from SeisBench, we perform systematic benchmark tests on data recorded by different types of instruments, ranging from acoustic emission sensors to distributed acoustic sensing, and collected at different scales, spanning from laboratory acoustic emission events to major tectonic earthquakes. Our tests highlight that rescaling is essential for dealing with small‐magnitude seismic events recorded at high sampling rates as well as larger magnitude events having long coda and remote events with long wave trains. Our results demonstrate that the proposed methods are effective in augmenting pre‐trained models for out‐of‐distribution data sets, especially in scenarios with limited labeled data for transfer learning.

Data versioning in machine learning is of paramount importance as it ensures the reproducibility, transparency, and reliability of ML models. In the dynamic landscape of ML research, where models heavily rely on diverse datasets, data versioning plays a crucial role in maintaining consistency throughout the ML pipeline. By tracking changes in datasets over time and aligning machine learning models with specific versions of data, researchers can reproduce experiments, verify results, and address challenges related to data quality, collaboration, and model training. Effective data versioning practices contribute to the robustness of ML workflows, fostering trust in model outcomes and supporting advancements in the field.

This project will formulate a machine learning based model using human genome data and forecast the chance for patients developing to severe dengue, a severe form of dengue fever and a particular condition that could undoubtedly end up being life hazardous. The research seeks to improve treatment by identifying genetic factors that predispose some people to a severe form of dengue and thus reduce mortality. Genomic data analysis and training of machine learning techniques such as feature selection, model training, and interpretability are incorporated in this study. Predicting severe dengue prognosis using ML models (SVM, random forest, deep learning models) for their training. Some tools like SHAP and LIME—create a predictive model for severe dengue prognosis. Collecting and preprocessing of genomic data, selection of the important genetic markers and validation of the performance of the model with accuracy and sensitivity metrics can be key steps. The final aim is to create a clinically handy product with a risk score for severe dengue and all other important genetic markers that can be used by health professionals in the future. Exploiting the existence of other cases and the past history to predict future cases, aim the proposed model to be more tailored and efficient in dengue cases management. By using genetic data t o understand better the risks of severe dengue, it will help to improve public health efforts and patient care.