Multiple Sclerosis (MS) is a chronic disease characterized by progressive or alternate impairment of neurological functions (motor, sensory, visual, and cognitive). Predicting disease progression with a probabilistic and time-dependent approach might help in suggesting interventions that can delay the progression of the disease. However, extracting informative knowledge from irregularly collected longitudinal data is difficult, and missing data pose significant challenges. MS progression is measured through the Expanded Disability Status Scale (EDSS), which quantifies and monitors disability in MS over time. EDSS assesses impairment in eight functional systems (FS). Frequently, only the EDSS score assigned by clinicians is reported, while FS sub-scores are missing. Imputing these scores might be useful, especially to stratify patients according to their phenotype assessed over the disease progression. This study aimed at i) exploring different methodologies for imputing missing FS sub-scores, and ii) predicting the EDSS score using complete clinical data. Results show that Exponential Weighted Moving Average achieved the lowest error rate in the missing data imputation task; furthermore, the combination of Classification and Regression Trees for the imputation and SVM for the prediction task obtained the best accuracy.

Clarity: The main paper is mostly written fairly well, the Appendix less so (lots of typos at least). The work nevertheless lacks clarity because several relevant details are moved to the Supplementary part, and some aspects are not mentioned at all (at least in the main paper). The Appendix even contains a section regarding categorical features that is not even hinted at in the main paper. Clarification is needed, e.g., at the following points: - p.2, l.70-73 is too vague, the meaning is unclear - pls. clarify - p.2, l. 85f: clarify what "[...] i enters leaf node l " means (i.e., that data pt. If \hat{y}_i denotes a predicted label, then why is it real-valued and not in [Y]? (Also regarding the description on p.3, l.96f: why should y_i - \hat{y}_i \geq 1 here -- \hat{y}_i is in R, so couldn't it be, say, y_i - delta for some small delta?) - p.3, l.92: perhaps clarify "tree sparsity" -- actually here this means sparsity of the decision hyperplanes, no the tree itself - The 1-norm is used in the MIP (1) and several times in the text later called "linear" (e.g., p.4, l.136), but this is technically incorrect.

This paper is about employing advances in computational efficiency of mixed integer programming methods towards decision tree construction problems. While locally optimal methods can achieve an upper bound on the minimization problem efficiently, closing the optimality gap requires tight lower bounds. The authors use an interval relaxation and a support-vector machine procedure to tighten the lower bound. To scale the algorithm, the authors use a LP-based data selection procedure, and perform all experiments using this procedure. It is not clear whether the global optimality properties of the MIP formulation carry through with the data-selection procedure.

The paper is well written, and its structure is adapted to the content. Upon reading the paper, one might think that the contribution resides in the vertical splitting of the data over the workers, but the state of the art study presented later on shows that this idea by itself is not new. The novelty comes from associating it with data also distributed vertically, sparse bit vectors for inter-node communications, feature compression with custom data structures and training on compressed data. The paper shows formally and experimentally how the proposed heuristics significantly improve the communication between the nodes and speed up training. The remark that using run-length encoding for the features allows them to hold in the L3 cache, thus decreasing the number of DRAM accesses, doesn't seem to always be true. The paper should explain in which conditions this is true (size of the cache, size of the data, number and type of features, etc.).

The idea of taking into account feature costs when pruning tree ensembles is original to the best of my knowledge. The main originality of the proposed approach is the fact that it adopts a bottom-up post-pruning strategy, while most existing approaches are top-down, acting during tree growing. While the authors present this feature as an advantage of their method, actually, I'm not convinced that adopting a bottom-up strategy is a good idea for addressing this problem. Since the algorithm indeed can not modify the existing tree structure (it can only prune it), it should be less efficient in terms of feature cost reduction than top-down methods that can have a direct impact on the features selected at tree nodes. For example, let us assume that two very important features in the dataset carry on the exact same information about the output (i.e, they are redundant).

Given the popularity of decision trees, proposing an efficient parallel implementation of this method is of course very relevant. The proposed parallelization is original with respect to existing methods and it should indeed lead to less communications than other methods. The theoretical analysis is sound and I like the discussion of the impact of the main problem and method parameters that follows from the lower bound provided in theorem 4.1. Experiments are conducted on two very large problems, where, in the limit of the tested settings (see below), PV-tree is clearly shown to outperform other parallel implementations, in terms of both computing times to reach a given accuracy level and communication costs. I nevertheless have two major concerns with the proposed parallelization.

Paper Summary: The paper presents a technique for testing whether a given set of samples are drawn from a postulated Gaussian latent tree model or a saturated Gaussian graphical model. The paper first characterizes a set of necessary and sufficient constraints that any covariance matrix of a Gaussian latent tree model should satisfy. It then uses these constraints to come up with a test statistic. The paper extends past work on testing for Gaussian latent tree models to settings where the observed variables are allowed to have degree up to 2. The test statistic presented in the paper is based on gaussian approximation for maxima of high dimensional sums. Simulations suggest that the test statistic can potentially work in high dimensional settings.

We propose a simple generalization of standard and empirically successful decision tree learning algorithms such as ID3, C4.5, and CART. These algorithms, which have been central to machine learning for decades, are greedy in nature: they grow a decision tree by iteratively splitting on the best attribute. Our algorithm, Top- k, considers the k best attributes as possible splits instead of just the single best attribute. We demonstrate, theoretically and empirically, the power of this simple generalization. We first prove a greediness hierarchy theorem showing that for every k\in \mathbb{N}, Top- (k 1) can be dramatically more powerful than Top- k: there are data distributions for which the former achieves accuracy 1-\epsilon, whereas the latter only achieves accuracy \frac{1}{2} \epsilon .

Decision trees are regarded for high interpretability arising from their hierarchical partitioning structure built on simple decision rules. However, in practice, this is not realized because axis-aligned partitioning of realistic data results in deep trees, and because ensemble methods are used to mitigate overfitting. Even then, model complexity and performance remain sensitive to transformation of the input, and extensive expert crafting of features from the raw data is common. We propose the first system to alternate sparse feature learning with differentiable decision tree construction to produce small, interpretable trees with good performance. We benchmark against conventional tree-based models and demonstrate several notions of interpretation of a model and its predictions.

The Landsat program is the longest-running Earth observation program in history, with 50 years of data acquisition by 8 satellites. The multispectral imagery captured by sensors onboard these satellites is critical for a wide range of scientific fields. Despite the increasing popularity of deep learning and remote sensing, the majority of researchers still use decision trees and random forests for Landsat image analysis due to the prevalence of small labeled datasets and lack of foundation models. In this paper, we introduce SSL4EO-L, the first ever dataset designed for Self-Supervised Learning for Earth Observation for the Landsat family of satellites (including 3 sensors and 2 product levels) and the largest Landsat dataset in history (5M image patches). Additionally, we modernize and re-release the L7 Irish and L8 Biome cloud detection datasets, and introduce the first ML benchmark datasets for Landsats 4–5 TM and Landsat 7 ETM SR.