Recent advances in machine learning have led to a surge in adoption of neural networks for various tasks, but lack of interpretability remains an issue for many others in which an understanding of the features influencing the prediction is necessary to ensure fairness, safety, and legal compliance. In this paper we consider one class of such tasks, tabular dataset classification, and propose a novel neuro-symbolic architecture, Neural Reasoning Networks (NRN), that is scalable and generates logically sound textual explanations for its predictions. NRNs are connected layers of logical neurons that implement a form of real valued logic. A training algorithm (R-NRN) learns the weights of the network as usual using gradient descent optimization with backprop, but also learns the network structure itself using a bandit-based optimization. Both are implemented in an extension to PyTorch that takes full advantage of GPU scaling and batched training. Evaluation on a diverse set of 22 open-source datasets for tabular classification demonstrates performance (measured by ROC AUC) which improves over Multilayer Perceptron (MLP) and is statistically similar to other state-of-the-art approaches such as Random Forest, XGBoost and Gradient Boosted Trees, while offering 43% faster training and a more than 2 orders of magnitude reduction in the number of parameters required, on average. Furthermore, R-NRN explanations are shorter than the compared approaches while producing more accurate feature importance scores.

Various forms of sparse attention have been explored to mitigate the quadratic computational and memory cost of the attention mechanism in transformers. We study sparse transformers not through a lens of efficiency but rather in terms of learnability and generalization. Empirically studying a range of attention mechanisms, we find that input-dependent sparse attention models appear to converge faster and generalize better than standard attention models, while input-agnostic sparse attention models show no such benefits -- a phenomenon that is robust across architectural and optimization hyperparameter choices. This can be interpreted as demonstrating that concentrating a model's "semantic focus" with respect to the tokens currently being considered (in the form of input-dependent sparse attention) accelerates learning. We develop a theoretical characterization of the conditions that explain this behavior. We establish a connection between the stability of the standard softmax and the loss function's Lipschitz properties, then show how sparsity affects the stability of the softmax and the subsequent convergence and generalization guarantees resulting from the attention mechanism. This allows us to theoretically establish that input-agnostic sparse attention does not provide any benefits. We also characterize conditions when semantic focus (input-dependent sparse attention) can provide improved guarantees, and we validate that these conditions are in fact met in our empirical evaluations.

Kinjal Basu, Keerthiram Murugesan, Subhajit Chaudhury, Murray Campbell, Kartik Talamadupula, Tim Klinger. Proceedings of the 18th Conference of the European Chapter of the Association for Computational Linguistics (Volume 1: Long Papers). 2024.

Maxwell Crouse, Pavan Kapanipathi, Subhajit Chaudhury, Tahira Naseem, Ramon Fernandez Astudillo, Achille Fokoue, Tim Klinger. Proceedings of the 61st Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers). 2023.

Humans have a remarkable ability to rapidly generalize to new tasks that is difficult to reproduce in artificial learning systems.Compositionality has been proposed as a key mechanism supporting generalization in humans, but evidence of its neural implementation and impact on behavior is still scarce. Here we study the computational properties associated with compositional generalization in both humans and artificial neural networks (ANNs) on a highly compositional task. First, we identified behavioral signatures of compositional generalization in humans, along with their neural correlates using whole-cortex functional magnetic resonance imaging (fMRI) data. Next, we designed pretraining paradigms aided by a procedure we term primitives pretraining to endow compositional task elements into ANNs. We found that ANNs with this prior knowledge had greater correspondence with human behavior and neural compositional signatures. Importantly, primitives pretraining induced abstract internal representations, excellent zero-shot generalization, and sample-efficient learning. Moreover, it gave rise to a hierarchy of abstract representations that matched human fMRI data, where sensory rule abstractions emerged in early sensory areas, and motor rule abstractions emerged in later motor areas. Our findings give empirical support to the role of compositional generalization in humans behavior, implicate abstract representations as its neural implementation, and illustrate that these representations can be embedded into ANNs by designing simple and efficient pretraining procedures.

Training agents to learn efficiently in multi-agent environments can benefit from the explicit modelling of other agent's beliefs, especially in complex limited-information games such as the Hanabi card game. However, generalization is also highly relevant to performance in these games, though model comparisons at large training timescales can be difficult. In this work, we address this by introducing a novel model trained using a sleep metaphor on a reduced complexity version of the Hanabi game. This sleep metaphor consists an altered training regiment, as well as an information-theoretic constraint on the agent's policy. Results from experimentation demonstrate improved performance through this sleep-metaphor method, and provide a promising motivation for using similar techniques in more complex methods that incorporate explicit models of other agent's beliefs.

This paper explores information-theoretic constraints on methods for multi-agent reinforcement learning (MARL) in mixed cooperative and competitive games. Within this domain, decentralized training has been employed to increase learning sample efficiency. However, these approaches do not explicitly discourage complex policies, which can lead to overfitting. To address this, we apply an information theoretic constraint onto agents' policies that discourages overly complex behaviour when it is not associated with a significant increase in reward. A second challenge in MARL is the non-stationarity of the environment introduced by other agents' changing policies. Previous methods in MARL have sought to reduce the impact of non-stationarity by inferring other agents' policies, but this can lead to over-fitting to previously observed behaviour. To avoid this, a similar information-theoretic constraint is applied onto the inference of other agents' policies, resulting in a more robust estimate. We evaluate the effects of these information-theoretic constraints on a test suite of multi-agent games, and report an overall improvement in performance, with greater improvements found in competitive domains compared to cooperative games.

The difficulty of deterministic planning increases exponentially with search-tree depth. Black-box planning presents an even greater challenge, since planners must operate without an explicit model of the domain. Heuristics can make search more efficient, but goal-aware heuristics for black-box planning usually rely on goal counting, which is often quite uninformative. In this work, we show how to overcome this limitation by discovering macro-actions that make the goal-count heuristic more accurate. Our approach searches for macro-actions with focused effects (i.e. macros that modify only a small number of state variables), which align well with the assumptions made by the goal-count heuristic. Focused macros dramatically improve black-box planning efficiency across a wide range of planning domains, sometimes beating even state-of-the-art planners with access to a full domain model.

Given the recent success of Deep Learning applied to a variety of single tasks, it is natural to consider more human-realistic settings. Perhaps the most difficult of these settings is that of continual lifelong learning, where the model must learn online over a continuous stream of non-stationary data. A successful continual lifelong learning system must have three key capabilities: it must learn and adapt over time, it must not forget what it has learned, and it must be efficient in both training time and memory. Recent techniques have focused their efforts primarily on the first two capabilities while questions of efficiency remain largely unexplored. In this paper, we consider the problem of efficient and effective storage of experiences over very large time-frames. In particular we consider the case where typical experiences are O(n) bits and memories are limited to O(k) bits for k

In recent years researchers have achieved considerable success applying neural network methods to question answering (QA). These approaches have achieved state of the art results in simplified closed-domain settings such as the SQuAD (Rajpurkar et al. 2016) dataset, which provides a pre-selected passage, from which the answer to a given question may be extracted. More recently, researchers have begun to tackle open-domain QA, in which the model is given a question and access to a large corpus (e.g., wikipedia) instead of a pre-selected passage (Chen et al. 2017a). This setting is more complex as it requires large-scale search for relevant passages by an information retrieval component, combined with a reading comprehension model that “reads” the passages to generate an answer to the question. Performance in this setting lags well behind closed-domain performance. In this paper, we present a novel open-domain QA system called Reinforced Ranker-Reader (R3), based on two algorithmic innovations. First, we propose a new pipeline for open-domain QA with a Ranker component, which learns to rank retrieved passages in terms of likelihood of extracting the ground-truth answer to a given question. Second, we propose a novel method that jointly trains the Ranker along with an answer-extraction Reader model, based on reinforcement learning. We report extensive experimental results showing that our method significantly improves on the state of the art for multiple open-domain QA datasets.

We introduce a new class of models called multiresolution recurrent neural networks, which explicitly model natural language generation at multiple levels of abstraction. The models extend the sequence-to-sequence framework to generate two parallel stochastic processes: a sequence of high-level coarse tokens, and a sequence of natural language words (e.g. sentences). The coarse sequences follow a latent stochastic process with a factorial representation, which helps the models generalize to new examples. The coarse sequences can also incorporate task-specific knowledge, when available. In our experiments, the coarse sequences are extracted using automatic procedures, which are designed to capture compositional structure and semantics. These procedures enable training the multiresolution recurrent neural networks by maximizing the exact joint log-likelihood over both sequences. We apply the models to dialogue response generation in the technical support domain and compare them with several competing models. The multiresolution recurrent neural networks outperform competing models by a substantial margin, achieving state-of-the-art results according to both a human evaluation study and automatic evaluation metrics. Furthermore, experiments show the proposed models generate more fluent, relevant and goal-oriented responses.

A Data Scientist typically performs a number of tedious and time-consuming steps to derive insight from a raw data set. The process usually starts with data ingestion, cleaning, and transformation (e.g. outlier removal, missing value imputation), then proceeds to model building, and finally a presentation of predictions that align with the end-users objectives and preferences. It is a long, complex, and sometimes artful process requiring substantial time and effort, especially because of the combinatorial explosion in choices of algorithms (and platforms), their parameters, and their compositions. Tools that can help automate steps in this process have the potential to accelerate the time-to-delivery of useful results, expand the reach of data science to non-experts, and offer a more systematic exploration of the available options. This work presents a step towards this goal.

Fixing defects is an essential software development activity. For commercial software vendors, the time to repair defects in deployed business-critical software products or applications is a key quality metric for sustained customer satisfaction. In this paper, we report on the analysis of about 1,500 defect records from an IBM middle-ware product collected over a five-year period. The analysis includes a characterization of each repaired defect by topic and a ranking of developers by inferred expertise on each topic. We find clear evidence that defect resolution time is strongly influenced by a specific developer and his/her expertise in the defect's topic. To validate our approach, we conducted interviews with the product's manager who provided us with his own ranking of developer expertise for comparison. We argue that our automated developer expertise ranking can be beneficial in the planning of a software project and is applicable beyond software support in the other phases of the software lifecycle.

We present a minimalist approach to model-based test generation. Our approach leverages the information provided in the domain and behavioral models of an application under test to produce a small yet effective set of test cases without additional user input. The domain model consists of UML class diagram with invariants, while the behavioral model consists of UML use cases. Each use case flow has an associated guard condition and a set of updates (to the domain object diagram and the output parameters). We treat the model invariants to enable a novel specify once, test everywhere paradigm. Our approach frees the modeler from the responsibility of specifying appropriate alternate flows on use cases affected by each invariant - our analysis augments the specified use cases with appropriate alternate flows. Our approach then produces suitable testing goals which are refinements of the guard conditions on the augmented flows using a set of fault models. Another salient feature of our approach is generation of verification sequences to ensure that the object diagram updates associated with a given flow are implemented correctly. Our technique uses a novel set of fault models to mutate an object diagram and a novel algorithm which distinguishes between the original and the mutated object diagrams. We describe the techniques used in our test generation approach.

The growing acceptance of model-based test generation in industry has created a need for a tester-friendly model construction process and associated tools. In this paper, we present an automated approach for checking consistency and completeness of testing models for interactive systems. The testing models are described in terms of operations provided by the system under test (SUT). Each operation is specified as a set of possible results each with a guard condition and a set of actions on its parameters and the system state. The consistency and completeness checks performed include identification of malformed models, irrelevant state variables, incomplete and redundant guard conditions, missing data values, and nondeterminism in and executability of operation results. We review our testing requirements model, and describe various consistency checks performed on this model. We illustrate these checks with an example test model, and also present algorithms used to perform our analysis