Understanding the function learned by a neural network is crucial in many domains, e.g., to detect a model’s adaption to concept drift in online learning. Existing global surrogate model approaches generate explanations by maximizing the fidelity between the neural network and a surrogate model on a sample-basis, which can be very time-consuming. Therefore, these approaches are not applicable in scenarios where timely or frequent explanations are required. In this paper, we introduce a real-time approach for generating a symbolic representation of the function learned by a neural network. Our idea is to generate explanations via another neural network (called the Interpretation Network, or -Net), which maps network parameters to a symbolic representation of the network function. We show that the training of an -Net for a family of functions can be performed up-front and subsequent generation of an explanation only requires querying the -Net once, which is computationally very efficient and does not require training data. We empirically evaluate our approach for the case of low-order polynomials as explanations, and show that it achieves competitive results for various data and function complexities. To the best of our knowledge, this is the first approach that attempts to learn mapping from neural networks to symbolic representations.
8 November 2023 in Journal of Geophysical Research: Earth Surface
During the RESOLVE project (“High‐resolution imaging in subsurface geophysics: development of a multi‐instrument platform for interdisciplinary research”), continuous surface displacement and seismic array observations were obtained on Glacier d’Argentière in the French Alps for 35 days in May 2018. The data set is used to perform a detailed study of targeted processes within the highly dynamic cryospheric environment. In particular, the physical processes controlling glacial basal motion are poorly understood and remain challenging to observe directly. Especially in the Alpine region for temperate based glaciers where the ice rapidly responds to changing climatic conditions and thus, processes are strongly intermittent in time and heterogeneous in space. Spatially dense seismic and Global Positioning System (GPS) measurements are analyzed applying machine learning to gain insight into the processes controlling glacial motions of Glacier d’Argentière. Using multiple bandpass‐filtered copies of the continuous seismic waveforms, we compute energy‐based features, develop a matched field beamforming catalog and include meteorological observations. Features describing the data are analyzed with a gradient boosting decision tree model to directly estimate the GPS displacements from the seismic noise. We posit that features of the seismic noise provide direct access to the dominant parameters that drive displacement on the highly variable and unsteady surface of the glacier. The machine learning model infers daily fluctuations and longer term trends. The results show on‐ice displacement rates are strongly modulated by activity at the base of the glacier. The techniques presented provide a new approach to study glacial basal sliding and discover its full complexity.
We consider generating explanations for neural networks in cases where the network’s training data is not accessible, for instance due to privacy or safety issues. Recently, Interpretation Nets ( $$\mathcal {I}$$ -Nets) have been proposed as a sample-free approach to post-hoc, global model interpretability that does not require access to training data. They formulate interpretation as a machine learning task that maps network representations (parameters) to a representation of an interpretable function. In this paper, we extend the $$\mathcal {I}$$ -Net framework to the cases of standard and soft decision trees as surrogate models. We propose a suitable decision tree representation and design of the corresponding $$\mathcal {I}$$ -Net output layers. Furthermore, we make $$\mathcal {I}$$ -Nets applicable to real-world tasks by considering more realistic distributions when generating the $$\mathcal {I}$$ -Net’s training data. We empirically evaluate our approach against traditional global, post-hoc interpretability approaches and show that it achieves superior results when the training data is not accessible.
Despite the success of deep learning for text and image data, tree-based ensemble models are still state-of-the-art for machine learning with heterogeneous tabular data. However, there is a significant need for tabular-specific gradient-based methods due to their high flexibility. In this paper, we propose $\text{GRANDE}$, $\text{GRA}$die$\text{N}$t-Based $\text{D}$ecision Tree $\text{E}$nsembles, a novel approach for learning hard, axis-aligned decision tree ensembles using end-to-end gradient descent. GRANDE is based on a dense representation of tree ensembles, which affords to use backpropagation with a straight-through operator to jointly optimize all model parameters. Our method combines axis-aligned splits, which is a useful inductive bias for tabular data, with the flexibility of gradient-based optimization. Furthermore, we introduce an advanced instance-wise weighting that facilitates learning representations for both, simple and complex relations, within a single model. We conducted an extensive evaluation on a predefined benchmark with 19 classification datasets and demonstrate that our method outperforms existing gradient-boosting and deep learning frameworks on most datasets. The method is available under: https://github.com/s-marton/GRANDE
Neural architectures such as Recurrent Neural Networks (RNNs), Transformers, and State-Space Models have shown great success in handling sequential data by learning temporal dependencies. Decision Trees (DTs), on the other hand, remain a widely used class of models for structured tabular data but are typically not designed to capture sequential patterns directly. Instead, DT-based approaches for time-series data often rely on feature engineering, such as manually incorporating lag features, which can be suboptimal for capturing complex temporal dependencies. To address this limitation, we introduce ReMeDe Trees, a novel recurrent DT architecture that integrates an internal memory mechanism, similar to RNNs, to learn long-term dependencies in sequential data. Our model learns hard, axis-aligned decision rules for both output generation and state updates, optimizing them efficiently via gradient descent. We provide a proof-of-concept study on synthetic benchmarks to demonstrate the effectiveness of our approach.