Recently, self‑supervised representation learning relying on vast amounts of unlabeled data has been explored as a pre‑training method for autonomous driving. However, directly applying popular contrastive or generative methods to this problem is insufficient and may even lead to negative transfer. In this paper, we present AD‑L‑JEPA, a novel self‑supervised pre‑training framework with a joint embedding predictive architecture (JEPA) for automotive LiDAR object detection. Unlike existing methods, AD‑L‑JEPA is neither generative nor contrastive. Instead of explicitly generating masked regions, our method predicts Bird’s‑Eye‑View embeddings to capture the diverse nature of driving scenes. Furthermore, our approach eliminates the need to manually form contrastive pairs by employing explicit variance regularization to avoid representation collapse. Experimental results demonstrate consistent improvements on the LiDAR 3D object detection downstream task across the KITTI3D, Waymo, and ONCE datasets, while reducing GPU hours by 1.9×–2.7× and GPU memory by 2.8×–4× compared with the state-of-the-art method Occupancy-MAE. Notably, on the largest ONCE dataset, pre‑training on 100K frames yields a 1.61 mAP gain, better than in case of all the other methods pre‑trained on either 100K or 500K frames, and pre‑training on 500K frames yields a 2.98 mAP gain, better than in case of all the other methods pre‑trained on either 500K or 1M frames. AD‑L‑JEPA constitutes the first JEPA‑based pre‑training method for autonomous driving. It offers better quality, faster, and more GPU‑memory‑efficient self‑supervised representation learning.

We study distributed training of deep learning models in time-constrained environments. We propose a new algorithm that periodically pulls workers towards the center variable computed as a weighted average of workers, where the weights are inversely proportional to the gradient norms of the workers such that recovering the flat regions in the optimization landscape is prioritized. We develop two asynchronous variants of the proposed algorithm that we call Model-level and Layer-level Gradient-based Weighted Averaging (resp. MGRAWA and LGRAWA), which differ in terms of the weighting scheme that is either done with respect to the entire model or is applied layer-wise. On the theoretical front, we prove the convergence guarantee for the proposed approach in both convex and non-convex settings. We then experimentally demonstrate that our algorithms outperform the competitor methods by achieving faster convergence and recovering better quality and flatter local optima. We also carry out an ablation study to analyze the scalability of the proposed algorithms in more crowded distributed training environments. Finally, we report that our approach requires less frequent communication and fewer distributed updates compared to the state-of-the-art baselines.

This work focuses on the decentralized deep learning optimization framework. We propose Adjacent Leader Decentralized Gradient Descent (AL-DSGD), for improving final model performance, accelerating convergence, and reducing the communication overhead of decentralized deep learning optimizers. AL-DSGD relies on two main ideas. Firstly, to increase the influence of the strongest learners on the learning system it assigns weights to different neighbor workers according to both their performance and the degree when averaging among them, and it applies a corrective force on the workers dictated by both the currently best-performing neighbor and the neighbor with the maximal degree. Secondly, to alleviate the problem of the deterioration of the convergence speed and performance of the nodes with lower degrees, AL-DSGD relies on dynamic communication graphs, which effectively allows the workers to communicate with more nodes while keeping the degrees of the nodes low. Experiments demonstrate that AL-DSGD accelerates the convergence of the decentralized state-of-the-art techniques and improves their test performance especially in the communication constrained environments. We also theoretically prove the convergence of the proposed scheme. Finally, we release to the community a highly general and concise PyTorch-based library for distributed training of deep learning models that supports easy implementation of any distributed deep learning approach ((a)synchronous, (de)centralized).

Automotive radar has been extensively utilized in cars for many years as an essential sensor, primarily due to its robustness in extreme weather conditions, its capacity to measure Doppler information in the surrounding environment, and its cost-effectiveness. Recently, developments in radar technologies and the availability of open-source radar data sets have attracted more attention to radars and using them for perception tasks in deep learning based autonomous driving. However, annotating radar data for large-scale autonomous driving perception tasks is challenging, i.e., it is difficult for humans to label this data and often requires a semi-automatic approach that involves projecting labels from other sensors, such as cameras and LiDARs. The lack of high-quality labeled data has limited the performance of radar perception models. In this paper, we propose MVRAE, a Multi-View Radar AutoEncoder, which employs self-supervised learning to learn meaningful representations from multi-view radar data without any labels. Our approach is based on the intuition that a good representation for multi-view radar data, which includes range-angle, range-Doppler, and angle-Doppler views, should enable the reconstruction of one view solely from the representations of the other two views. Experimental results demonstrate that our proposed self-supervised method, that can be used as a pre-training step for autonomous driving task, allows the model to learn meaningful representations from unlabeled radar data and achieves enhanced label efficiency for downstream tasks, such as radar semantic segmentation. To the best of our knowledge, MVRAE is the first work that employs self-supervised learning and conducts systematic experiments with multi-view radar data.

Modern deep learning (DL) architectures are trained using variants of the SGD algorithm and typically rely on the user to manually drop the learning rate when the training curve saturates. In this paper, we develop an algorithm, that we call AutoDrop, that realizes the learning rate drop automatically and stems from the properties of the learning dynamics of DL systems. Specifically, it is motivated by the observation that the angular velocity of the model parameters, i.e., the velocity of the changes of the convergence direction, for a fixed learning rate initially increases rapidly and then progresses towards soft saturation. At saturation, the optimizer slows down thus the angular velocity saturation is a good indicator for dropping the learning rate. After the drop, the angular velocity {“}resets{”} and follows the pattern described above, increasing again until saturation. AutoDrop is built on this idea and drops the learning rate whenever the angular velocity saturates. The method is simple to implement, computationally cheap, and by design avoids the short-horizon bias problem. We show that AutoDrop achieves favorable performance compared to many different baseline manual and automatic learning rate schedulers, and matches the SOTA performance on all our experiments. On the theoretical front, we claim two contributions: we formulate the learning rate behavior based on the angular velocity and provide general convergence theory for the learning rate schedulers that decrease the learning rate step-wise, rather than continuously as is commonly analyzed.

Among various sensors for assisted and autonomous driving systems, automotive radar has been considered as a robust and low-cost solution even in adverse weather or lighting conditions. With the recent development of radar technologies and open-sourced annotated data sets, semantic segmentation with radar signals has become very promising. However, existing methods are either computationally expensive or discard significant amounts of valuable information from raw 3D radar signals by reducing them to 2D planes via averaging. In this work, we introduce ERASE-Net, an Efficient RAdar SEgmentation Network to segment the raw radar signals semantically. The core of our approach is the novel detect-then-segment method for raw radar signals. It first detects the center point of each object, then extracts a compact radar signal representation, and finally performs semantic segmentation. We show that our method can achieve superior performance on radar semantic segmentation task compared to the state-of-the-art (SOTA) technique. Furthermore, our approach requires up to 20×less computational resources. Finally, we show that the proposed ERASE-Net can be compressed by 40% without significant loss in performance, significantly more than the SOTA network, which makes it a more promising candidate for practical automotive applications.

Interpretability is crucial to understand the inner workings of deep neural networks (DNNs). Many interpretation methods help to understand the decision-making of DNNs by generating saliency maps that highlight parts of the input image that contribute the most to the prediction made by the DNN. In this paper we design a backdoor attack that alters the saliency map produced by the network for an input image with a specific trigger pattern while not losing the prediction performance significantly. The saliency maps are incorporated in the penalty term of the objective function that is used to train a deep model and its influence on model training is conditioned upon the presence of a trigger. We design two types of attacks: a targeted attack that enforces a specific modification of the saliency map and a non-targeted attack when the importance scores of the top pixels from the original saliency map are significantly reduced. We perform empirical evaluations of the proposed backdoor attacks on gradient-based interpretation methods, Grad-CAM and SimpleGrad, and a gradient-free scheme, VisualBackProp, for a variety of deep learning architectures. We show that our attacks constitute a serious security threat to the reliability of the interpretation methods when deploying models developed by untrusted sources. We furthermore show that existing backdoor defense mechanisms are ineffective in detecting our attacks. Finally, we demonstrate that the proposed methodology can be used in an inverted setting, where the correct saliency map can be obtained only in the presence of a trigger (key), effectively making the interpretation system available only to selected users.

In this paper, we study the sharpness of a deep learning (DL) loss landscape around local minima in order to reveal systematic mechanisms underlying the generalization abilities of DL models. Our analysis is performed across varying network and optimizer hyper-parameters, and involves a rich family of different sharpness measures. We compare these measures and show that the low-pass filter based measure exhibits the highest correlation with the generalization abilities of DL models, has high robustness to both data and label noise, and furthermore can track the double descent behavior for neural networks. We next derive the optimization algorithm, relying on the low-pass filter (LPF), that actively searches the flat regions in the DL optimization landscape using SGD-like procedure. The update of the proposed algorithm, that we call LPF-SGD, is determined by the gradient of the convolution of the filter kernel with the loss function and can be efficiently computed using MC sampling. We empirically show that our algorithm achieves superior generalization performance compared to the common DL training strategies. On the theoretical front we prove that LPF-SGD converges to a better optimal point with smaller generalization error than SGD.

We consider multi-label classification where the goal is to annotate each data point with the most relevant subset of labels from an extremely large label set. Efficient annotation can be achieved with balanced tree predictors, i.e. trees with logarithmic-depth in the label complexity, whose leaves correspond to labels. Designing prediction mechanism with such trees for real data applications is non-trivial as it needs to accommodate sending examples to multiple leaves while at the same time sustain high prediction accuracy. In this paper we develop the LdSM algorithm for the construction and training of multi-label decision trees, where in every node of the tree we optimize a novel objective function that favors balanced splits, maintains high class purity of children nodes, and allows sending examples to multiple directions but with a penalty that prevents tree over-growth. Each node of the tree is trained once the previous node is completed leading to a streaming approach for training. We analyze the proposed objective theoretically and show that minimizing it leads to pure and balanced data splits. Furthermore, we show a boosting theorem that captures its connection to the multi-label classification error. Experimental results on benchmark data sets demonstrate that our approach achieves high prediction accuracy and low prediction time and position LdSM as a competitive tool among existing state-of-the-art approaches.

We introduce a new approach for comparing reinforcement learning policies, using Wasserstein distances (WDs) in a newly defined latent behavioral space. We show that by utilizing the dual formulation of the WD, we can learn score functions over policy behaviors that can in turn be used to lead policy optimization towards (or away from) (un)desired behaviors. Combined with smoothed WDs, the dual formulation allows us to devise efficient algorithms that take stochastic gradient descent steps through WD regularizers. We incorporate these regularizers into two novel on-policy algorithms, Behavior-Guided Policy Gradient and Behavior-Guided Evolution Strategies, which we demonstrate can outperform existing methods in a variety of challenging environments. We also provide an open source demo.

End-to-end learning from sensory data has shown promising results in autonomous driving. While employing many sensors enhances world perception and should lead to more robust and reliable behavior of autonomous vehicles, it is challenging to train and deploy such network and at least two problems are encountered in the considered setting. The first one is the increase of computational complexity with the number of sensing devices. The other is the phenomena of network overfitting to the simplest and most informative input. We address both challenges with a novel, carefully tailored multi-modal experts network architecture and propose a multi-stage training procedure. The network contains a gating mechanism, which selects the most relevant input at each inference time step using a mixed discrete-continuous policy. We demonstrate the plausibility of the proposed approach on our 1/6 scale truck equipped with three cameras and one LiDAR.

Despite significant recent advances in deep neural networks, training them remains a challenge due to the highly non-convex nature of the objective function. State-of-the-art methods rely on error backpropagation, which suffers from several well-known issues, such as vanishing and exploding gradients, inability to handle non-differentiable nonlinearities and to parallelize weight-updates across layers, and biological implausibility. These limitations continue to motivate exploration of alternative training algorithms, including several recently proposed auxiliary-variable methods which break the complex nested objective function into local subproblems. However, those techniques are mainly offline (batch), which limits their applicability to extremely large datasets, as well as to online, continual or reinforcement learning. The main contribution of our work is a novel online (stochastic/mini-batch) alternating minimization (AM) approach for training deep neural networks, together with the first theoretical convergence guarantees for AM in stochastic settings and promising empirical results on a variety of architectures and datasets.

We propose a reconfigurable network for efficient inference dedicated to autonomous platforms equipped with multiple perception sensors. The size of the network for steering autonomous platforms grows proportionally to the number of installed sensors eventually preventing the usage of multiple sensors in real-time applications due to an inefficient inference. Our approach hinges on the observation that multiple sensors provide a large stream of data, where only a fraction of the data is relevant for the performed task at any given moment in time. The architecture of the reconfigurable network that we propose contains separate feature extractors, called experts, for each sensor. The decisive block of our model is the gating network, which online decides which sensor provides the data that is most relevant for driving. It then reconfigures the network by activating only the relevant expert corresponding to that sensor and deactivating the remaining ones. As a consequence, the model never extracts features from data that are irrelevant for driving. The gating network takes the data from all inputs and thus to avoid explosion of computation time and memory space it has to be realized as a small and shallow network. We verify our model on the unmanned ground vehicle (UGV) comprising of the 1/6 scale remote control truck equipped with three cameras. We demonstrate that the reconfigurable network correctly chooses experts in real-time allowing the reduction of computations cost for the whole model without deteriorating its performance.

We consider distributed optimization under communication constraints for training deep learning models. We propose a new algorithm, whose parameter updates rely on two forces: a regular gradient step, and a corrective direction dictated by the currently best-performing worker (leader). Our method differs from the parameter-averaging scheme EASGD in a number of ways: (i) our objective formulation does not change the location of stationary points compared to the original optimization problem; (ii) we avoid convergence decelerations caused by pulling local workers descending to different local minima to each other (i.e. to the average of their parameters); (iii) our update by design breaks the curse of symmetry (the phenomenon of being trapped in poorly generalizing sub-optimal solutions in symmetric non-convex landscapes); and (iv) our approach is more communication efficient since it broadcasts only parameters of the leader rather than all workers. We provide theoretical analysis of the batch version of the proposed algorithm, which we call Leader Gradient Descent (LGD), and its stochastic variant (LSGD). Finally, we implement an asynchronous version of our algorithm and extend it to the multi-leader setting, where we form groups of workers, each represented by its own local leader (the best performer in a group), and update each worker with a corrective direction comprised of two attractive forces: one to the local, and one to the global leader (the best performer among all workers). The multi-leader setting is well-aligned with current hardware architecture, where local workers forming a group lie within a single computational node and different groups correspond to different nodes. For training convolutional neural networks, we empirically demonstrate that our approach compares favorably to state-of-the-art baselines.

This paper proposes a new method, that we call VisualBackProp, for visualizing which sets of pixels of the input image contribute most to the predictions made by the convolutional neural network (CNN). The method heavily hinges on exploring the intuition that the feature maps contain less and less irrelevant information to the prediction decision when moving deeper into the network. The technique we propose is dedicated for CNN-based systems for steering self-driving cars and is therefore required to run in real-time. This makes the proposed visualization method a valuable debugging tool which can be easily used during both training and inference. We justify our approach with theoretical arguments and confirm that the proposed method identifies sets of input pixels, rather than individual pixels, that collaboratively contribute to the prediction. We utilize the proposed visualization tool in the NVIDIA neural-network-based end-to-end learning system for autonomous driving, known as PilotNet. We demonstrate that VisualBackProp determines which elements in the road image most influence PilotNet's steering decision and indeed captures relevant objects on the road. The empirical evaluation furthermore shows the plausibility of the proposed approach on public road video data as well as in other applications and reveals that it compares favorably to the layer-wise relevance propagation approach, i.e. it obtains similar visualization results and achieves order of magnitude speed-ups.

The paper proposes an on-line monitoring framework for continuous real-time safety/security in learning-based control systems (specifically application to a unmanned ground vehicle). We monitor validity of mappings from sensor inputs to actuator commands, controller-focused anomaly detection (CFAM), and from actuator commands to sensor inputs, system-focused anomaly detection (SFAM). CFAM is an image conditioned energy based generative adversarial network (EBGAN) in which the energy based discriminator distinguishes between proper and anomalous actuator commands. SFAM is based on an action condition video prediction framework to detect anomalies between predicted and observed temporal evolution of sensor data. We demonstrate the effectiveness of the approach on our autonomous ground vehicle for indoor environments and on Udacity dataset for outdoor environments.

We consider an efficient computational framework for speeding up several machine learning algorithms with almost no loss of accuracy. The proposed framework relies on projections via structured matrices that we call Structured Spinners, which are formed as products of three structured matrix-blocks that incorporate rotations. The approach is highly generic, i.e. i) structured matrices under consideration can either be fully-randomized or learned, ii) our structured family contains as special cases all previously considered structured schemes, iii) the setting extends to the non-linear case where the projections are followed by non-linear functions, and iv) the method finds numerous applications including kernel approximations via random feature maps, dimensionality reduction algorithms,new fast cross-polytope LSH techniques, deep learning, convex optimization algorithms via Newton sketches, quantization with random projection trees, and more. The proposed framework comes with theoretical guarantees characterizing the capacity of the structured model in reference to its unstructured counterpart and is based on a general theoretical principle that we describe in the paper. As a consequence of our theoretical analysis, we provide the first theoretical guarantees for one of the most efficient existing LSH algorithms based on the HD 3 HD 2 HD 1 structured matrix [Andoni et al., 2015]. The exhaustive experimental evaluation confirms the accuracy and efficiency of structured spinners for a variety of different applications.

We consider multi-class classification where the predictor has a hierarchical structure that allows for a very large number of labels both at train and test time. The predictive power of such models can heavily depend on the structure of the tree, and although past work showed how to learn the tree structure, it expected that the feature vectors remained static. We provide a novel algorithm to simultaneously perform representation learning for the input data and learning of the hierarchical predictor. Our approach optimizes an objective function which favors balanced and easily-separable multi-way node partitions. We theoretically analyze this objective, showing that it gives rise to a boosting style property and a bound on classification error. We next show how to extend the algorithm to conditional density estimation. We empirically validate both variants of the algorithm on text classification and language modeling, respectively, and show that they compare favorably to common baselines in terms of accuracy and running time.

We present a novel end-to-end learning framework to enable ground vehicles to autonomously navigate unknown environments by fusing raw pixels from cameras and depth measurements from a LiDAR. A deep neural network architecture is introduced to effectively perform modality fusion and reliably predict steering commands even in the presence of sensor failures. The proposed network is trained on our own dataset, from LiDAR and a camera mounted on a UGV taken in an indoor corridor environment. Comprehensive experimental evaluation to demonstrate the robustness of our network architecture is performed to show that the proposed deep learning neural network is able to autonomously navigate in the corridor environment. Furthermore, we demonstrate that the fusion of the camera and LiDAR modalities provides further benefits beyond robustness to sensor failures. Specifically, the multimodal fused system shows a potential to navigate around static and dynamic obstacles and to handle changes in environment geometry without being trained for these tasks.

We consider the hashing mechanism for constructing binary embeddings, that involves pseudo-random projections followed by nonlinear (sign function) mappings. The pseudo-random projection is described by a matrix, where not all entries are independent random variables but instead a fixed “budget of randomness” is distributed across the matrix. Such matrices can be edfficiently stored in sub-quadratic or even linear space, provide reduction in randomness usage (i.e. number of required random values), and very often lead to computational speed ups. We prove several theoretical results showing that projections via various structured matrices followed by nonlinear mappings accurately preserve the angular distance between input high-dimensional vectors. To the best of our knowledge, these results are the first that give theoretical ground for the use of general structured matrices in the nonlinear setting. We empirically verify our theoretical findings and show the dependence of learning via structured hashed projections on the performance of neural network as well as nearest neighbor classifier.

We study the connection between the highly non-convex loss function of a simple model of the fully-connected feed-forward neural network and the Hamiltonian of the spherical spin-glass model under the assumptions of: i) variable independence, ii) redundancy in network parametrization, and iii) uniformity. These assumptions enable us to explain the complexity of the fully decoupled neural network through the prism of the results from random matrix theory. We show that for large-size decoupled networks the lowest critical values of the random loss function form a layered structure and they are located in a well-defined band lower-bounded by the global minimum. The number of local minima outside that band diminishes exponentially with the size of the network. We empirically verify that the mathematical model exhibits similar behavior as the computer simulations, despite the presence of high dependencies in real networks. We conjecture that both simulated annealing and SGD converge to the band of low critical points, and that all critical points found there are local minima of high quality measured by the test error. This emphasizes a major difference between large- and small-size networks where for the latter poor quality local minima have non-zero probability of being recovered. Finally, we prove that recovering the global minimum becomes harder as the network size increases and that it is in practice irrelevant as global minimum often leads to overfitting.

Deep learning has enjoyed a resurgence of interest in the last few years for such applications as image and speech recognition, or natural language processing. The vast majority of practical applications of deep learning focus on supervised learning, where the supervised loss function is minimized using stochastic gradient descent. The properties of this highly non-convex loss function, such as its landscape and the behavior of critical points (maxima, minima, and saddle points), as well as the reason why large- and small-size networks achieve radically different practical performance, are however very poorly understood. It was only recently shown that new results in spin-glass theory potentially may provide an explanation for these problems by establishing a connection between the loss function of the neural networks and the Hamiltonian of the spherical spin-glass models. The connection between both models relies on a number of possibly unrealistic assumptions, yet the empirical evidence suggests that the connection may exist in real. The question we pose is whether it is possible to drop some of these assumptions to establish a stronger connection between both models.

We study the problem of multiclass classification with an extremely large number of classes (k), with the goal of obtaining train and test time complexity logarithmic in the number of classes. We develop top-down tree construction approaches for constructing logarithmic depth trees. On the theoretical front, we formulate a new objective function, which is optimized at each node of the tree and creates dynamic partitions of the data which are both pure (in terms of class labels) and balanced. We demonstrate that under favorable conditions, we can construct logarithmic depth trees that have leaves with low label entropy. However, the objective function at the nodes is challenging to optimize computationally. We address the empirical problem with a new online decision tree construction procedure. Experiments demonstrate that this online algorithm quickly achieves improvement in test error compared to more common logarithmic training time approaches, which makes it a plausible method in computationally constrained large-k applications.

We study the problem of stochastic optimization for deep learning in the parallel computing environment under communication constraints. A new algorithm is proposed in this setting where the communication and coordination of work among concurrent processes (local workers), is based on an elastic force which links the parameters they compute with a center variable stored by the parameter server (master). The algorithm enables the local workers to perform more exploration, i.e. the algorithm allows the local variables to fluctuate further from the center variable by reducing the amount of communication between local workers and the master. We empirically demonstrate that in the deep learning setting, due to the existence of many local optima, allowing more exploration can lead to the improved performance. We propose synchronous and asynchronous variants of the new algorithm. We provide the stability analysis of the asynchronous variant in the round-robin scheme and compare it with the more common parallelized method ADMM. We show that the stability of EASGD is guaranteed when a simple stability condition is satisfied, which is not the case for ADMM. We additionally propose the momentum-based version of our algorithm that can be applied in both synchronous and asynchronous settings. Asynchronous variant of the algorithm is applied to train convolutional neural networks for image classification on the CIFAR and ImageNet datasets. Experiments demonstrate that the new algorithm accelerates the training of deep architectures compared to DOWNPOUR and other common baseline approaches and furthermore is very communication efficient.

The partition function plays a key role in probabilistic modeling including conditional random fields, graphical models, and maximum likelihood estimation. To optimize partition functions, this article introduces a quadratic variational upper bound. This inequality facilitates majorization methods: optimization of complicated functions through the iterative solution of simpler sub-problems. Such bounds remain efficient to compute even when the partition function involves a graphical model (with small tree-width) or in latent likelihood settings. For large-scale problems, low-rank versions of the bound are provided and outperform LBFGS as well as first-order methods. Several learning applications are shown and reduce to fast and convergent update rules. Experimental results show advantages over state-of-the-art optimization methods.