Textual Explanations and Critiques in Recommendation Systems

Artificial intelligence and machine learning algorithms have become ubiquitous. Although they offer a wide range of benefits, their adoption in decision-critical fields is limited by their lack of interpretability, particularly with textual data. Moreover, with more data available than ever before, it has become increasingly important to explain automated predictions.Generally, users find it difficult to understand the underlying computational processes and interact with the models, especially when the models fail to generate the outcomes or explanations, or both, correctly. This problem highlights the growing need for users to better understand the models' inner workings and gain control over their actions. This dissertation focuses on two fundamental challenges of addressing this need. The first involves explanation generation: inferring high-quality explanations from text documents in a scalable and data-driven manner. The second challenge consists in making explanations actionable, and we refer to it as critiquing. This dissertation examines two important applications in natural language processing and recommendation tasks.First, we present two models for extracting high-quality and personalized justifications from text documents under the paradigm of selective rationalization - a binary selection of input features from which the model computes the outcome. We propose the concept of multi-dimensional rationales and emphasize that a single overall selection does not accurately capture the multi-faceted nature of useful rationales. The first method relies on multi-task learning and learns the rationales in an unsupervised manner, with no prior in the data. Inspired by the role of concept-based thinking in human reasoning, our second model is a generalization: it assumes that only one label is observed. We empirically demonstrate the efficiency of the models in terms of explanation and predictive performance as well as their scalability. It opens new doors for applications, such as recommendation or summarization.Second, we use the high-level idea of multi-dimensional rationalization to mine massive corpora and construct large personalized justification datasets. We show that human users significantly prefer our explanations over those produced by state-of-the-art methods.Last, we propose two conversational explainable recommendation systems. The first explains a user rating by inferring a set of keyphrases. Conditioned on them, it generates an abstractive personal justification. We allow users to interact iteratively with the explanations and refine the recommendation through an unsupervised critiquing method. Our second model is based on multimodal modeling and self-supervision. It enables fast and efficient multi-step critiquing. Using real-world datasets, we show that our models exhibit better performance than state-of-the-art models in terms of recommendation, explanation, and critiquing.Overall, we demonstrate that interpretability does not come at the cost of reduced performance in two consequential applications. Our framework is applicable to other fields as well. This dissertation presents an effective means of closing the gap between promise and practice in artificial intelligence.

Lifelong Machine Learning with Data Efficiency and Knowledge Retention

Fei Mi

Artificial intelligence (AI) and machine learning (ML) have become de facto tools in many real-life applications to offer a wide range of benefits for individuals and our society. A classic ML model is typically trained with a large-scale static dataset in an offline manner. Therefore, it can not quickly capture new knowledge in non-stationary environments, and it is difficult to maintain long-term memory for knowledge learned earlier. In practice, many ML systems often need to learn new knowledge (e.g., domains, tasks, distributions, etc.) as more data and experiences are collected, which is referred to as a lifelong ML paradigm in this thesis. We focus on two fundamental challenges to achieve lifelong learning. The first challenge is to quickly learn new knowledge with a small number of observations, and we refer to it as data efficiency. The second challenge is to prevent an ML system from forgetting the old knowledge it has previously learned, and we refer to this challenge as knowledge retention. These two capabilities are crucial for applying ML to most practical applications. In this thesis, we study three important applications with these two challenges, including recommendation systems, task-oriented dialog systems, and the image classification task. First, we propose two approaches to improve data efficiency for task-oriented dialog systems. The first proposed approach is based on Meta-learning, aiming to learn a better model parameter initialization from training data. It can quickly reach a good parameter region of new domains or tasks with a small number of labeled data. The second proposal takes a semi-supervised self-training approach to iteratively train a better model using sufficient unlabeled data when only a limited number of labeled data are available. We empirically demonstrate that both approaches effectively improve data efficiency to learn new knowledge. The second self-training method even consistently improves state-of-the-art large-scale pre-trained models. Second, we tackle the knowledge retention challenge to mitigate the detrimental catastrophic forgetting issue when neural networks learn new knowledge sequentially. We formulate and investigate the ``continual learning'' setting for task-oriented dialog systems and recommendation systems. Through extensive empirical evaluation and analysis, we demonstrate the importance of (1) exemplar replay: storing representative historical data and replaying them to the model while learning new knowledge; (2) dynamic regularization: applying a dynamic regularization term to put flexible constraints on not forgetting previously learned knowledge in each model update cycle. Lastly, we conduct several initial attempts to achieve both data efficiency and knowledge retention in a unified framework. In the recommendation scenario, we propose two approaches using different non-parametric memory modules to retain long-term knowledge. More importantly, the two proposed non-parametric predictions computed on top of them help learn and memorize new knowledge in a data-efficient manner. Apart from the recommendation scenario, we propose a probabilistic evaluation protocol in the widely studied image classification domain. It is general and versatile to simulate a wide range of realistic lifelong learning scenarios that require both knowledge retention and data efficiency for studying different techniques. Through experiments, we also demonstrate the benefit

EPFL2021

Prior knowledge in Kernel methods

Alexei Pozdnoukhov

Machine Learning is a modern and actively developing field of computer science, devoted to extracting and estimating dependencies from empirical data. It combines such fields as statistics, optimization theory and artificial intelligence. In practical tasks, the general aim of Machine Learning is to construct algorithms able to generalize and predict in previously unseen situations based on some set of examples. Given some finite information, Machine Learning provides ways to exract knowledge, describe, explain and predict from data. Kernel Methods are one of the most successful branches of Machine Learning. They allow applying linear algorithms with well-founded properties such as generalization ability, to non-linear real-life problems. Support Vector Machine is a well-known example of a kernel method, which has found a wide range of applications in data analysis nowadays. In many practical applications, some additional prior knowledge is often available. This can be the knowledge about the data domain, invariant transformations, inner geometrical structures in data, some properties of the underlying process, etc. If used smartly, this information can provide significant improvement to any data processing algorithm. Thus, it is important to develop methods for incorporating prior knowledge into data-dependent models. The main objective of this thesis is to investigate approaches towards learning with kernel methods using prior knowledge. Invariant learning with kernel methods is considered in more details. In the first part of the thesis, kernels are developed which incorporate prior knowledge on invariant transformations. They apply when the desired transformation produce an object around every example, assuming that all points in the given object share the same class. Different types of objects, including hard geometrical objects and distributions are considered. These kernels were then applied for images classification with Support Vector Machines. Next, algorithms which specifically include prior knowledge are considered. An algorithm which linearly classifies distributions by their domain was developed. It is constructed such that it allows to apply kernels to solve non-linear tasks. Thus, it combines the discriminative power of support vector machines and the well-developed framework of generative models. It can be applied to a number of real-life tasks which include data represented as distributions. In the last part of the thesis, the use of unlabelled data as a source of prior knowledge is considered. The technique of modelling the unlabelled data with a graph is taken as a baseline from semi-supervised manifold learning. For classification problems, we use this apporach for building graph models of invariant manifolds. For regression problems, we use unlabelled data to take into account the inner geometry of the input space. To conclude, in this thesis we developed a number of approaches for incorporating some prior knowledge into kernel methods. We proposed invariant kernels for existing algorithms, developed new algorithms and adapted a technique taken from semi-supervised learning for invariant learning. In all these cases, links with related state-of-the-art approaches were investigated. Several illustrative experiments were carried out on real data on optical character recognition, face image classification, brain-computer interfaces, and a number of benchmark and synthetic datasets.

EPFL2006

System Support for Robust Distributed Learning

Arsany Hany Abdelmessih Guirguis

Machine learning (ML) applications are ubiquitous. They run in different environments such as datacenters, the cloud, and even on edge devices. Despite where they run, distributing ML training seems the only way to attain scalable, high-quality learning. But, distributing ML is challenging, essentially due to the unique nature of ML applications.First, ML training needs to be robust against arbitrary (i.e., Byzantine) failures due to its usage in mission-critical applications. Second, training applications in datacenters run on shared clusters of computing resources, for which we need resource allocation solutions that meet the high computation demands of these applications while fully utilizing existing resources. Third, running distributed training in the cloud faces a network bottleneck, exacerbated by the fast-growing pace of computing power. Hence, we need solutions that reduce the communication load without impacting the training accuracy. Fourth, despite the scalability and privacy guarantees of training on edge devices via federated learning, the heterogeneity of devices' capabilities and their data distributions calls for robust solutions that cope with these challenges.To achieve robustness, we introduce Garfield, a library to help practitioners make their ML applications Byzantine-resilient. Besides addressing the vulnerability of the shared-graph architecture followed by classical ML frameworks, Garfield supports various communication patterns, robust aggregation rules, and compute hardware (i.e., CPUs and GPUs). We show how to use Garfield in different architectures, network settings, and data distributions.We explore elastic training (i.e., changing the training parameters mid-execution) to efficiently solve the resource allocation problem in datacenters' shared clusters. We present ERA, which provides elasticity in two dimensions: (1) it scales jobs horizontally, i.e., by adding or removing resources to or from the running jobs, and (2) it dynamically changes, at will, the per-GPU batch size to control the utilization of each GPU. We demonstrate that simultaneous scaling in both dimensions improves the training time without impacting the training accuracy.We show how to use cloud object stores (COS) to alleviate the network bottleneck of training transfer learning (TL) applications in the cloud. We propose HAPI, a processing system for TL that spans the compute and the COS tiers, enabling significant improvements while remaining transparent to the user. HAPI mitigates the network bottleneck by carefully splitting the TL application such that feature extraction is, partially or entirely, executed next to storage.We show how to efficiently and robustly train generative adversarial networks (GANs) in the federated learning paradigm with FeGAN. Essentially, we co-locate both components of a GAN (i.e., a generator and a discriminator) on each device (addressing the scaling problem) and have a server aggregate the devices' models using balanced sampling and Kullback-Leibler weighting, mitigating training issues and boosting convergence.

EPFL2022