Machine Learning (cs.LG)

Abstract: How to effectively represent molecules is a long-standing challenge for molecular property prediction and drug discovery. This paper studies this problem and proposes to incorporate chemical domain knowledge, specifically related to chemical reactions, for learning effective molecular representations. However, the inherent cross-modality property between chemical reactions and molecules presents a significant challenge to address. To this end, we introduce a novel method, namely MolKD, which Distills cross-modal Knowledge in chemical reactions to assist Molecular property prediction. Specifically, the reaction-to-molecule distillation model within MolKD transfers cross-modal knowledge from a pre-trained teacher network learning with one modality (i.e., reactions) into a student network learning with another modality (i.e., molecules). Moreover, MolKD learns effective molecular representations by incorporating reaction yields to measure transformation efficiency of the reactant-product pair when pre-training on reactions. Extensive experiments demonstrate that MolKD significantly outperforms various competitive baseline models, e.g., 2.1% absolute AUC-ROC gain on Tox21. Further investigations demonstrate that pre-trained molecular representations in MolKD can distinguish chemically reasonable molecular similarities, which enables molecular property prediction with high robustness and interpretability.

Abstract: Formal verification of neural networks is essential before their deployment in safety-critical settings. However, existing methods for formally verifying neural networks are not yet scalable enough to handle practical problems that involve a large number of neurons. In this work, we propose a novel approach to address this challenge: A conservative neural network reduction approach that ensures that the verification of the reduced network implies the verification of the original network. Our approach constructs the reduction on-the-fly, while simultaneously verifying the original network and its specifications. The reduction merges all neurons of a nonlinear layer with similar outputs and is applicable to neural networks with any type of activation function such as ReLU, sigmoid, and tanh. Our evaluation shows that our approach can reduce a network to less than 5% of the number of neurons and thus to a similar degree the verification time is reduced.

Abstract: The flexibility and effectiveness of message passing based graph neural networks (GNNs) induced considerable advances in deep learning on graph-structured data. In such approaches, GNNs recursively update node representations based on their neighbors and they gain expressivity through the use of node and edge attribute vectors. E.g., in computational tasks such as physics and chemistry usage of edge attributes such as relative position or distance proved to be essential. In this work, we address not what kind of attributes to use, but how to condition on this information to improve model performance. We consider three types of conditioning; weak, strong, and pure, which respectively relate to concatenation-based conditioning, gating, and transformations that are causally dependent on the attributes. This categorization provides a unifying viewpoint on different classes of GNNs, from separable convolutions to various forms of message passing networks. We provide an empirical study on the effect of conditioning methods in several tasks in computational chemistry.

Abstract: This paper aims to prove the emergence of symbolic concepts in well-trained AI models. We prove that if (1) the high-order derivatives of the model output w.r.t. the input variables are all zero, (2) the AI model can be used on occluded samples and will yield higher confidence when the input sample is less occluded, and (3) the confidence of the AI model does not significantly degrade on occluded samples, then the AI model will encode sparse interactive concepts. Each interactive concept represents an interaction between a specific set of input variables, and has a certain numerical effect on the inference score of the model. Specifically, it is proved that the inference score of the model can always be represented as the sum of the interaction effects of all interactive concepts. In fact, we hope to prove that conditions for the emergence of symbolic concepts are quite common. It means that for most AI models, we can usually use a small number of interactive concepts to mimic the model outputs on any arbitrarily masked samples.

Abstract: Dataset distillation is attracting more attention in machine learning as training sets continue to grow and the cost of training state-of-the-art models becomes increasingly high. By synthesizing datasets with high information density, dataset distillation offers a range of potential applications, including support for continual learning, neural architecture search, and privacy protection. Despite recent advances, we lack a holistic understanding of the approaches and applications. Our survey aims to bridge this gap by first proposing a taxonomy of dataset distillation, characterizing existing approaches, and then systematically reviewing the data modalities, and related applications. In addition, we summarize the challenges and discuss future directions for this field of research.

Abstract: Targeted model poisoning attacks pose a significant threat to federated learning systems. Recent studies show that edge-case targeted attacks, which target a small fraction of the input space are nearly impossible to counter using existing fixed defense strategies. In this paper, we strive to design a learned-defense strategy against such attacks, using a small defense dataset. The defense dataset can be collected by the central authority of the federated learning task, and should contain a mix of poisoned and clean examples. The proposed framework, LearnDefend, estimates the probability of a client update being malicious. The examples in defense dataset need not be pre-marked as poisoned or clean. We also learn a poisoned data detector model which can be used to mark each example in the defense dataset as clean or poisoned. We estimate the poisoned data detector and the client importance models in a coupled optimization approach. Our experiments demonstrate that LearnDefend is capable of defending against state-of-the-art attacks where existing fixed defense strategies fail. We also show that LearnDefend is robust to size and noise in the marking of clean examples in the defense dataset.

Abstract: Active flow control (AFC) involves manipulating fluid flow over time to achieve a desired performance or efficiency. AFC, as a sequential optimisation task, can benefit from utilising Reinforcement Learning (RL) for dynamic optimisation. In this work, we introduce Gym-preCICE, a Python adapter fully compliant with Gymnasium (formerly known as OpenAI Gym) API to facilitate designing and developing RL environments for single- and multi-physics AFC applications. In an actor-environment setting, Gym-preCICE takes advantage of preCICE, an open-source coupling library for partitioned multi-physics simulations, to handle information exchange between a controller (actor) and an AFC simulation environment. The developed framework results in a seamless non-invasive integration of realistic physics-based simulation toolboxes with RL algorithms. Gym-preCICE provides a framework for designing RL environments to model AFC tasks, as well as a playground for applying RL algorithms in various AFC-related engineering applications.

Abstract: Deep Reinforcement Learning agents often suffer from catastrophic forgetting, forgetting previously found solutions in parts of the input space when training on new data. Replay Memories are a common solution to the problem, decorrelating and shuffling old and new training samples. They naively store state transitions as they come in, without regard for redundancy. We introduce a novel cognitive-inspired replay memory approach based on the Grow-When-Required (GWR) self-organizing network, which resembles a map-based mental model of the world. Our approach organizes stored transitions into a concise environment-model-like network of state-nodes and transition-edges, merging similar samples to reduce the memory size and increase pair-wise distance among samples, which increases the relevancy of each sample. Overall, our paper shows that map-based experience replay allows for significant memory reduction with only small performance decreases.

Abstract: Constructing decision trees online is a classical machine learning problem. Existing works often assume that features are readily available for each incoming data point. However, in many real world applications, both feature values and the labels are unknown a priori and can only be obtained at a cost. For example, in medical diagnosis, doctors have to choose which tests to perform (i.e., making costly feature queries) on a patient in order to make a diagnosis decision (i.e., predicting labels). We provide a fresh perspective to tackle this practical challenge. Our framework consists of an active planning oracle embedded in an online learning scheme for which we investigate several information acquisition functions. Specifically, we employ a surrogate information acquisition function based on adaptive submodularity to actively query feature values with a minimal cost, while using a posterior sampling scheme to maintain a low regret for online prediction. We demonstrate the efficiency and effectiveness of our framework via extensive experiments on various real-world datasets. Our framework also naturally adapts to the challenging setting of online learning with concept drift and is shown to be competitive with baseline models while being more flexible.

Abstract: Robust loss functions are designed to combat the adverse impacts of label noise, whose robustness is typically supported by theoretical bounds agnostic to the training dynamics. However, these bounds may fail to characterize the empirical performance as it remains unclear why robust loss functions can underfit. We show that most loss functions can be rewritten into a form with the same class-score margin and different sample-weighting functions. The resulting curriculum view provides a straightforward analysis of the training dynamics, which helps attribute underfitting to diminished average sample weights and noise robustness to larger weights for clean samples. We show that simple fixes to the curriculums can make underfitting robust loss functions competitive with the state-of-the-art, and training schedules can substantially affect the noise robustness even with robust loss functions. Code is available at \url{github}.

Abstract: Sliced-Wasserstein Flow (SWF) is a promising approach to nonparametric generative modeling but has not been widely adopted due to its suboptimal generative quality and lack of conditional modeling capabilities. In this work, we make two major contributions to bridging this gap. First, based on a pleasant observation that (under certain conditions) the SWF of joint distributions coincides with those of conditional distributions, we propose Conditional Sliced-Wasserstein Flow (CSWF), a simple yet effective extension of SWF that enables nonparametric conditional modeling. Second, we introduce appropriate inductive biases of images into SWF with two techniques inspired by local connectivity and multiscale representation in vision research, which greatly improve the efficiency and quality of modeling images. With all the improvements, we achieve generative performance comparable with many deep parametric generative models on both conditional and unconditional tasks in a purely nonparametric fashion, demonstrating its great potential.

Abstract: Lottery Ticket Hypothesis (LTH) claims the existence of a winning ticket (i.e., a properly pruned sub-network together with original weight initialization) that can achieve competitive performance to the original dense network. A recent work, called UGS, extended LTH to prune graph neural networks (GNNs) for effectively accelerating GNN inference. UGS simultaneously prunes the graph adjacency matrix and the model weights using the same masking mechanism, but since the roles of the graph adjacency matrix and the weight matrices are very different, we find that their sparsifications lead to different performance characteristics. Specifically, we find that the performance of a sparsified GNN degrades significantly when the graph sparsity goes beyond a certain extent. Therefore, we propose two techniques to improve GNN performance when the graph sparsity is high. First, UGS prunes the adjacency matrix using a loss formulation which, however, does not properly involve all elements of the adjacency matrix; in contrast, we add a new auxiliary loss head to better guide the edge pruning by involving the entire adjacency matrix. Second, by regarding unfavorable graph sparsification as adversarial data perturbations, we formulate the pruning process as a min-max optimization problem to gain the robustness of lottery tickets when the graph sparsity is high. We further investigate the question: Can the "retrainable" winning ticket of a GNN be also effective for graph transferring learning? We call it the transferable graph lottery ticket (GLT) hypothesis. Extensive experiments were conducted which demonstrate the superiority of our proposed sparsification method over UGS, and which empirically verified our transferable GLT hypothesis.

Abstract: Data in many real-world applications are often accumulated over time, like a stream. In contrast to conventional machine learning studies that focus on learning from a given training data set, learning from data streams cannot ignore the fact that the incoming data stream can be potentially endless with overwhelming size and unknown changes, and it is impractical to assume to have sufficient computational/storage resource such that all received data can be handled in time. Thus, the generalization performance of learning from data streams depends not only on how many data have been received, but also on how many data can be well exploited timely, with resource and rapidity concerns, in addition to the ability of learning algorithm and complexity of the problem. For this purpose, in this article we introduce the notion of machine learning throughput, define Stream Efficient Learning and present a preliminary theoretical framework.

Abstract: We propose LESS-VFL, a communication-efficient feature selection method for distributed systems with vertically partitioned data. We consider a system of a server and several parties with local datasets that share a sample ID space but have different feature sets. The parties wish to collaboratively train a model for a prediction task. As part of the training, the parties wish to remove unimportant features in the system to improve generalization, efficiency, and explainability. In LESS-VFL, after a short pre-training period, the server optimizes its part of the global model to determine the relevant outputs from party models. This information is shared with the parties to then allow local feature selection without communication. We analytically prove that LESS-VFL removes spurious features from model training. We provide extensive empirical evidence that LESS-VFL can achieve high accuracy and remove spurious features at a fraction of the communication cost of other feature selection approaches.

Abstract: We establish matching upper and lower generalization error bounds for mini-batch Gradient Descent (GD) training with either deterministic or stochastic, data-independent, but otherwise arbitrary batch selection rules. We consider smooth Lipschitz-convex/nonconvex/strongly-convex loss functions, and show that classical upper bounds for Stochastic GD (SGD) also hold verbatim for such arbitrary nonadaptive batch schedules, including all deterministic ones. Further, for convex and strongly-convex losses we prove matching lower bounds directly on the generalization error uniform over the aforementioned class of batch schedules, showing that all such batch schedules generalize optimally. Lastly, for smooth (non-Lipschitz) nonconvex losses, we show that full-batch (deterministic) GD is essentially optimal, among all possible batch schedules within the considered class, including all stochastic ones.

Abstract: We develop and analyze a general technique for learning with an unknown distribution drift. Given a sequence of independent observations from the last $T$ steps of a drifting distribution, our algorithm agnostically learns a family of functions with respect to the current distribution at time $T$. Unlike previous work, our technique does not require prior knowledge about the magnitude of the drift. Instead, the algorithm adapts to the sample data. Without explicitly estimating the drift, the algorithm learns a family of functions with almost the same error as a learning algorithm that knows the magnitude of the drift in advance. Furthermore, since our algorithm adapts to the data, it can guarantee a better learning error than an algorithm that relies on loose bounds on the drift.

Abstract: During the continuous evolution of one organism's ancestry, its genes accumulate extensive experiences and knowledge, enabling newborn descendants to rapidly adapt to their specific environments. Motivated by this observation, we propose a novel machine learning paradigm \textit{Learngene} to enable learning models to incorporate three key characteristics of genes. (i) Accumulating: the knowledge is accumulated during the continuous learning of an \textbf{ancestry model}. (ii) Condensing: the exhaustive accumulated knowledge is condensed into a much more compact information piece, \ie \textbf{learngene}. (iii): Inheriting: the condensed \textbf{learngene} is inherited to make it easier for \textbf{descendant models} to adapt to new environments. Since accumulating has been studied in some well-developed paradigms like large-scale pre-training and lifelong learning, we focus on condensing and inheriting, which induces three key issues and we provide the preliminary solutions to these issues in this paper: (i) \textit{Learngene} Form: the \textbf{learngene} is set to a few integral layers that can preserve the most commonality. (ii) \textit{Learngene} Condensing: we identify which layers among the ancestry model have the most similarity as one pseudo descendant model. (iii) \textit{Learngene} Inheriting: to construct distinct descendant models for specific downstream tasks, we stack some randomly initialized layers to the \textbf{learngene} layers. Extensive experiments of various settings, including using different network architectures like Vision Transformer (ViT) and Convolutional Neural Networks (CNNs) on different datasets, are carried out to confirm five advantages and two characteristics of \textit{Learngene}.

Abstract: DST methods achieve state-of-the-art results in sparse neural network training, matching the generalization of dense models while enabling sparse training and inference. Although the resulting models are highly sparse and theoretically cheaper to train, achieving speedups with unstructured sparsity on real-world hardware is challenging. In this work we propose a DST method to learn a variant of structured N:M sparsity, the acceleration of which in general is commonly supported in commodity hardware. Furthermore, we motivate with both a theoretical analysis and empirical results, the generalization performance of our specific N:M sparsity (constant fan-in), present a condensed representation with a reduced parameter and memory footprint, and demonstrate reduced inference time compared to dense models with a naive PyTorch CPU implementation of the condensed representation Our source code is available at https://github.com/calgaryml/condensed-sparsity

Abstract: Large language models (LLMs) have demonstrated remarkable abilities in representation learning for program synthesis and understanding tasks. The quality of the learned representations appears to be dictated by the neural scaling laws as a function of the number of model parameters and observations, while imposing upper bounds on the model performance by the amount of available data and compute, which is costly. In this study, we attempt to render the training of LLMs for program synthesis more efficient by unifying four key components: (1) model architectures, (2) learning methods, (3) infill sampling, and, (4) data distributions. Specifically, for the model architecture, we attempt to unify encoder and decoder-based models into a single prefix-LM. For learning methods, (i) causal language modeling, (ii) span corruption, (iii) infilling are unified into a simple learning algorithm. For infill sampling, we explore the claim of a "free lunch" hypothesis. For data distributions, the effect of a mixture distribution of programming and natural languages on model performance is explored. We conduct a comprehensive series of empirical experiments on 1B LLMs, for which failures and successes of this exploration are distilled into four lessons. We will provide a final recipe for training and release CodeGen2 models in size 1B, 3.7B, 7B, and, 16B parameters, along with the training framework as open-source: https://github.com/salesforce/CodeGen2.