The support vector machine (SVM), developed by the computer science community in the 1990s, is a supervised learning algorithm commonly used and originally intended for a binary classification setting. It is often considered one of the best "out of the box" classifiers. The SVM is a generalization of the simple yet elegant algorithm called the maximal margin classifier. This classifier, however, cannot be applied in every situation since it relies heavily on the assumption that the dataset is linearly separable -- thus, several extensions exist. Note: In the following, we will only cover the maximal margin classifier, purposely avoiding the different extensions.

This post deals with a recent optimizing method for training neural networks described in the paper Sharpness-Aware Minimization for Efficiently Improving Generalization by P. Foret et al. (December 2020). Honestly, the first time I read about the paper details, I really thought the procedure therein described (or something similar) had already been explored many years before by tons of people… I was even surprised to read that it worked in some contexts. Modern models train through optimization methods relying just on the training loss. These models can easily memorize the training data and are prone to overfitting. They have more parameters than needed and this large number of parameters provides no guarantee of proper generalization to the test set.

On top of recommending the excellent autobiography of Stanislaw Ulam, this post is about using the software Stan, but not directly to perform inference, instead to obtain R functions to evaluate a target's probability density function and its gradient. With which, one can implement custom methods, while still benefiting from the great work of the Stan team on the "modeling language" side. As a proof of concept I have implemented a plain Hamiltonian Monte Carlo sampler for a random effect logistic regression model (taken from a course on Multilevel Models by Germán Rodríguez), a coupling of that HMC algorithm (as in "Unbiased Hamiltonian Monte Carlo with couplings", see also this very recent article on the topic of coupling HMC), and then upper bounds on the total variation distance between the chain and its limiting distribution, as in "Estimating Convergence of Markov chains with L-Lag Couplings". Basically the R script starts like a standard script that would use rstan for inference; it runs the default algorithm of Stan for a little while, then extracts some info from the "stanfit" object. With these, a pure R implementation of TV upper bounds for a naive HMC algorithm follows, that relies on functions called "stan_logtarget" and "stan_gradlogtarget" to evaluate the target log-pdf and its gradient.

Human activity recognition has grown in popularity with its increase of applications within daily lifestyles and medical environments. The goal of having efficient and reliable human activity recognition brings benefits such as accessible use and better allocation of resources; especially in the medical industry. Activity recognition and classification can be obtained using many sophisticated data recording setups, but there is also a need in observing how performance varies among models that are strictly limited to using sensor data from easily accessible devices: smartphones and smartwatches. This paper presents the findings of different models that are limited to train using such sensors. The models are trained using either the k-Nearest Neighbor, Support Vector Machine, or Random Forest classifier algorithms. Performance and evaluations are done by comparing various model performances using different combinations of mobile sensors and how they affect recognitive performances of models. Results show promise for models trained strictly using limited sensor data collected from only smartphones and smartwatches coupled with traditional machine learning concepts and algorithms.

Various unsupervised greedy selection methods have been proposed as computationally tractable approximations to the NP-hard subset selection problem. These methods rely on sequentially selecting the variables that best improve performance with respect to a selection criterion. Theoretical results exist that provide performance bounds and enable "lazy greedy" efficient implementations for selection criteria that satisfy a diminishing returns property known as submodularity. This has motivated the development of variable selection algorithms based on mutual information and frame potential. Recently, the authors introduced Forward Selection Component Analysis (FSCA) which uses variance explained as its selection criterion. While this criterion is not submodular, FSCA has been shown to be highly effective for applications such as measurement plan optimisation. In this paper a "lazy" implementation of the FSCA algorithm (L-FSCA) is proposed, which, although not equivalent to FSCA due to the absence of submodularity, has the potential to yield comparable performance while being up to an order of magnitude faster to compute. The efficacy of L-FSCA is demonstrated by performing a systematic comparison with FSCA and five other unsupervised variable selection methods from the literature using simulated and real-world case studies. Experimental results confirm that L-FSCA yields almost identical performance to FSCA while reducing computation time by between 22% and 94% for the case studies considered.

Link prediction aims to infer the existence of a link between two nodes in a network. Despite their wide application, the success of traditional link prediction algorithms is hindered by three major challenges -- link sparsity, node attribute noise and network dynamics -- that are faced by real-world networks. To overcome these challenges, we propose a Contextualized Self-Supervised Learning (CSSL) framework that fully exploits structural context prediction for link prediction. The proposed CSSL framework forms edge embeddings through aggregating pairs of node embeddings constructed via a transformation on node attributes, which are used to predict the link existence probability. To generate node embeddings tailored for link prediction, structural context prediction is leveraged as a self-supervised learning task to boost link prediction. Two types of structural contexts are investigated, i.e., context nodes collected from random walks vs. context subgraphs. The CSSL framework can be trained in an end-to-end manner, with the learning of node and edge embeddings supervised by link prediction and the self-supervised learning task. The proposed CSSL is a generic and flexible framework in the sense that it can handle both transductive and inductive link prediction settings, and both attributed and non-attributed networks. Extensive experiments and ablation studies on seven real-world benchmark graph datasets demonstrate the superior performance of the proposed self-supervision based link prediction algorithm over state-of-the-art baselines on different types of networks under both transductive and inductive settings. The proposed CSSL also yields competitive performance in terms of its robustness to node attribute noise and scalability over large-scale networks.

Entity alignment aims at integrating heterogeneous knowledge from different knowledge graphs. Recent studies employ embedding-based methods by first learning the representation of Knowledge Graphs and then performing entity alignment via measuring the similarity between entity embeddings. However, they failed to make good use of the relation semantic information due to the trade-off problem caused by the different objectives of learning knowledge embedding and neighborhood consensus. To address this problem, we propose Relational Knowledge Distillation for Entity Alignment (RKDEA), a Graph Convolutional Network (GCN) based model equipped with knowledge distillation for entity alignment. We adopt GCN-based models to learn the representation of entities by considering the graph structure and incorporating the relation semantic information into GCN via knowledge distillation. Then, we introduce a novel adaptive mechanism to transfer relational knowledge so as to jointly learn entity embedding and neighborhood consensus. Experimental results on several benchmarking datasets demonstrate the effectiveness of our proposed model.

In image generation, generative models can be evaluated naturally by visually inspecting model outputs. However, this is not always the case for graph generative models (GGMs), making their evaluation challenging. Currently, the standard process for evaluating GGMs suffers from three critical limitations: i) it does not produce a single score which makes model selection challenging, ii) in many cases it fails to consider underlying edge and node features, and iii) it is prohibitively slow to perform. In this work, we mitigate these issues by searching for scalar, domain-agnostic, and scalable metrics for evaluating and ranking GGMs. To this end, we study existing GGM metrics and neural-network-based metrics emerging from generative models of images that use embeddings extracted from a task-specific network. Motivated by the power of certain Graph Neural Networks (GNNs) to extract meaningful graph representations without any training, we introduce several metrics based on the features extracted by an untrained random GNN. We design experiments to thoroughly test metrics on their ability to measure the diversity and fidelity of generated graphs, as well as their sample and computational efficiency. Depending on the quantity of samples, we recommend one of two random-GNN-based metrics that we show to be more expressive than pre-existing metrics. While we focus on applying these metrics to GGM evaluation, in practice this enables the ability to easily compute the dissimilarity between any two sets of graphs regardless of domain. Our code is released at: https://github.com/uoguelph-mlrg/GGM-metrics.

Learning the embeddings for urban regions from human mobility data can reveal the functionality of regions, and then enables the correlated but distinct tasks such as crime prediction. Human mobility data contains rich but abundant information, which yields to the comprehensive region embeddings for cross domain tasks. In this paper, we propose multi-graph fusion networks (MGFN) to enable the cross domain prediction tasks. First, we integrate the graphs with spatio-temporal similarity as mobility patterns through a mobility graph fusion module. Then, in the mobility pattern joint learning module, we design the multi-level cross-attention mechanism to learn the comprehensive embeddings from multiple mobility patterns based on intra-pattern and inter-pattern messages. Finally, we conduct extensive experiments on real-world urban datasets. Experimental results demonstrate that the proposed MGFN outperforms the state-of-the-art methods by up to 12.35% improvement.

Automated Machine Learning (AutoML) has been used successfully in settings where the learning task is assumed to be static. In many real-world scenarios, however, the data distribution will evolve over time, and it is yet to be shown whether AutoML techniques can effectively design online pipelines in dynamic environments. This study aims to automate pipeline design for online learning while continuously adapting to data drift. For this purpose, we design an adaptive Online Automated Machine Learning (OAML) system, searching the complete pipeline configuration space of online learners, including preprocessing algorithms and ensembling techniques. This system combines the inherent adaptation capabilities of online learners with the fast automated pipeline (re)optimization capabilities of AutoML. Focusing on optimization techniques that can adapt to evolving objectives, we evaluate asynchronous genetic programming and asynchronous successive halving to optimize these pipelines continually. We experiment on real and artificial data streams with varying types of concept drift to test the performance and adaptation capabilities of the proposed system. The results confirm the utility of OAML over popular online learning algorithms and underscore the benefits of continuous pipeline redesign in the presence of data drift.