Hyperparameter tunning and why it matters Cross-validation and nested cross-validation Hyperparameter tunning with Grid and Random search Bayesian Optimisation Tree-Structured Parzen Estimators, Population Based Training and SMAC Hyperparameter tunning tools, i.e., Hyperopt, Optuna, Scikit-optimize, Keras Turner and others Welcome to Hyperparameter Optimization for Machine Learning. In this course, you will learn multiple techniques to select the best hyperparameters and improve the performance of your machine learning models. If you are regularly training machine learning models as a hobby or for your organization and want to improve the performance of your models, if you are keen to jump up in the leader board of a data science competition, or you simply want to learn more about how to tune hyperparameters of machine learning models, this course will show you how. We'll take you step-by-step through engaging video tutorials and teach you everything you need to know about hyperparameter tuning. Throughout this comprehensive course, we cover almost every available approach to optimize hyperparameters, discussing their rationale, their advantages and shortcomings, the considerations to have when using the technique and their implementation in Python.

As machine learning (ML) techniques are being increasingly used in many applications, their vulnerability to adversarial attacks becomes well-known. Test time attacks, usually launched by adding adversarial noise to test instances, have been shown effective against the deployed ML models. In practice, one test input may be leveraged by different ML models. Test time attacks targeting a single ML model often neglect their impact on other ML models. In this work, we empirically demonstrate that naively attacking the classifier learning one concept may negatively impact classifiers trained to learn other concepts. For example, for the online image classification scenario, when the Gender classifier is under attack, the (wearing) Glasses classifier is simultaneously attacked with the accuracy dropped from 98.69 to 88.42. This raises an interesting question: is it possible to attack one set of classifiers without impacting the other set that uses the same test instance? Answers to the above research question have interesting implications for protecting privacy against ML model misuse. Attacking ML models that pose unnecessary risks of privacy invasion can be an important tool for protecting individuals from harmful privacy exploitation. In this paper, we address the above research question by developing novel attack techniques that can simultaneously attack one set of ML models while preserving the accuracy of the other. In the case of linear classifiers, we provide a theoretical framework for finding an optimal solution to generate such adversarial examples. Using this theoretical framework, we develop a multi-concept attack strategy in the context of deep learning. Our results demonstrate that our techniques can successfully attack the target classes while protecting the protected classes in many different settings, which is not possible with the existing test-time attack-single strategies.

State-of-the-art deep learning models have achieved significant performance levels on various benchmarks. However, the excellent performance comes at a cost of inefficient computational cost. Light-weight architectures, on the other hand, achieve moderate accuracies, but at a much more desirable latency. This paper presents a new method of jointly using the large accurate models together with the small fast ones. To this end, we propose an Energy-Based Joint Reasoning (EBJR) framework that adaptively distributes the samples between shallow and deep models to achieve an accuracy close to the deep model, but latency close to the shallow one. Our method is applicable to out-of-the-box pre-trained models as it does not require an architecture change nor re-training. Moreover, it is easy to use and deploy, especially for cloud services. Through a comprehensive set of experiments on different down-stream tasks, we show that our method outperforms strong state-of-the-art approaches with a considerable margin. In addition, we propose specialized EBJR, an extension of our method where we create a smaller specialized side model that performs the target task only partially, but yields an even higher accuracy and faster inference. We verify the strengths of our methods with both theoretical and experimental evaluations.

Pervasive computing promotes the installation of connected devices in our living spaces in order to provide services. Two major developments have gained significant momentum recently: an advanced use of edge resources and the integration of machine learning techniques for engineering applications. This evolution raises major challenges, in particular related to the appropriate distribution of computing elements along an edge-to-cloud continuum. About this, Federated Learning has been recently proposed for distributed model training in the edge. The principle of this approach is to aggregate models learned on distributed clients in order to obtain a new, more general model. The resulting model is then redistributed to clients for further training. To date, the most popular federated learning algorithm uses coordinate-wise averaging of the model parameters for aggregation. However, it has been shown that this method is not adapted in heterogeneous environments where data is not identically and independently distributed (non-iid). This corresponds directly to some pervasive computing scenarios where heterogeneity of devices and users challenges machine learning with the double objective of generalization and personalization. In this paper, we propose a novel aggregation algorithm, termed FedDist, which is able to modify its model architecture (here, deep neural network) by identifying dissimilarities between specific neurons amongst the clients. This permits to account for clients' specificity without impairing generalization. Furthermore, we define a complete method to evaluate federated learning in a realistic way taking generalization and personalization into account. Using this method, FedDist is extensively tested and compared with three state-of-the-art federated learning algorithms on the pervasive domain of Human Activity Recognition with smartphones.

If the authors of a recent Stanford report (Bommasani et al., 2021) on the opportunities and risks of "foundation models" are to be believed, these models represent a paradigm shift for AI and for the domains in which they will supposedly be used, including education. Although the name is new (and contested (Field, 2021)), the term describes existing types of algorithmic models that are "trained on broad data at scale" and "fine-tuned" (i.e., adapted) for particular downstream tasks, and is intended to encompass large language models such as BERT or GPT-3 and computer vision models such as CLIP. Such technologies have the potential for harm broadly speaking (e.g., Bender et al., 2021), but their use in the educational domain is particularly fraught, despite the potential benefits for learners claimed by the authors. In section 3.3 of the Stanford report, Malik et al. argue that achieving the goal of providing education for all learners requires more efficient computational approaches that can rapidly scale across educational domains and across educational contexts, for which they argue foundation models are uniquely well-suited. However, evidence suggests that not only are foundation models not likely to achieve the stated benefits for learners, but their use may also introduce new risks for harm.

The principle of independent causal mechanisms (ICM) states that generative processes of real world data consist of independent modules which do not influence or inform each other. While this idea has led to fruitful developments in the field of causal inference, it is not widely-known in the NLP community. In this work, we argue that the causal direction of the data collection process bears nontrivial implications that can explain a number of published NLP findings, such as differences in semi-supervised learning (SSL) and domain adaptation (DA) performance across different settings. We categorize common NLP tasks according to their causal direction and empirically assay the validity of the ICM principle for text data using minimum description length. We conduct an extensive meta-analysis of over 100 published SSL and 30 DA studies, and find that the results are consistent with our expectations based on causal insights. This work presents the first attempt to analyze the ICM principle in NLP, and provides constructive suggestions for future modeling choices. Code available at https://github.com/zhijing-jin/icm4nlp

Machine learning algorithms have become prevalent tools for decision-making in socially sensitive situations, such as determining credit-score ratings or predicting recidivism during parole. It has been recognized that algorithms are capable of learning societal biases, for example with respect to race (Larson, Mattu, Kirchner, and Angwin 2016) or gender (Lambrecht and Tucker 2019; Blau and Kahn 2003), and this realization seeded an important debate in the machine learning community about fairness of algorithms and their impact on decision-making. In order to define and measure discrimination, existing intuitive notions have been statistically formalized, thereby providing fairness metrics. For example, demographic parity (Darlington 1971) requires the protected attribute A (gender/race/religion etc.) to be independent of a constructed classifier or regressor Ŷ, written as Ŷ A. Another notion, termed equality of odds (Hardt, Price, Srebro et al. 2016), requires equal false positive and false negative rates of classifier Ŷ between different groups (females and males for example), written as Ŷ A Y. To this day, various different notions of fairness exist, which are sometimes incompatible (Corbett-Davies and Goel 2018), meaning not of all of them can be achieved for a predictor Ŷ simultaneously. There is still no consensus on which notion of fairness is the correct one. The discussion on algorithmic fairness is, however, not restricted to the machine learning domain. There are many legal and philosophical aspects that have arisen. For example, the legal distinction between disparate impact and disparate treatment (McGinley 2011) is important for assessing fairness from a judicial point of view.

The paper considers the problem of multi-objective decision support when outcomes are uncertain. We extend the concept of Pareto-efficient decisions to take into account the uncertainty of decision outcomes across varying contexts. This enables quantifying trade-offs between decisions in terms of tail outcomes that are relevant in safety-critical applications. We propose a method for learning efficient decisions with statistical confidence, building on results from the conformal prediction literature. The method adapts to weak or nonexistent context covariate overlap and its statistical guarantees are evaluated using both synthetic and real data.

The course will start off at the basics and work all the way through the financial accounting topics generally covered in an undergraduate program. First, we will describe what financial accounting is and the objectives of financial accounting. We will learn how the double-entry accounting system works by applying it to the accounting equation. In other words, we will use an accounting equation to record financial transactions using a double-entry accounting system. We well learn all topics by fist having presentations and then applying the skills using Excel practice problems.

In the past decade, machine learning has become a familiar technology for improving the efficiency and accuracy of processes like recommendations, supply chain forecasting, developing chatbots, image and text search, and automated customer service functions, to name a few. Machine learning today is becoming even more pervasive, impacting every market segment and industry, including manufacturing, SaaS platforms, health care, reservations and customer support routing, natural language processing (NLP) tasks such as intelligent document processing, and even food services. Take the case of Domino's Pizza, which has been using machine learning tools created to improve efficiencies in pizza production. "Domino's had a project called Project 3/10, which aimed to have a pizza ready for pickup within three minutes of an order, or have it delivered within 10 minutes of an order," says Dr. Bratin Saha, vice president and general manager of machine learning services for Amazon AI. "If you want to hit those goals, you have to be able to predict when a pizza order will come in. They use predictive machine learning models to achieve that." The recent rise of machine learning across diverse industries has been driven by improvements in other technological areas, says Saha--not the least of which is the increasing compute power in cloud data centers. "Over the last few years," explains Saha, "the amount of total compute that can be thrown at machine learning problems has been doubling almost every four months. As a result, a lot of functions that once could only be done by humans--things like detecting an object or understanding speech--are being performed by computers and machine learning models." "At AWS, everything we do works back from the customer and figuring out how we reduce their pain points and how we make it easier for them to do machine learning. At the bottom of the stack of machine learning services, we are innovating on the machine learning infrastructure so that we can make it cheaper for customers to do machine learning and faster for customers to do machine learning. There we have two AWS innovations. One is Inferentia and the other is Trainium."