One fundamental challenge in graph machine learning is generalizing to new graphs. Many existing methods following the inductive setup can generalize to test graphs with new structures, but assuming the feature and label spaces remain the same as the training ones. This paper introduces a fully-inductive setup, where models should perform inference on arbitrary test graphs with new structures, feature and label spaces. We propose GraphAny as the first attempt at this challenging setup. GraphAny models inference on a new graph as an analytical solution to a LinearGNN, which can be naturally applied to graphs with any feature and label spaces. To further build a stronger model with learning capacity, we fuse multiple LinearGNN predictions with learned inductive attention scores. Specifically, the attention module is carefully parameterized as a function of the entropy-normalized distance features between pairs of LinearGNN predictions to ensure generalization to new graphs. Empirically, GraphAny trained on a single Wisconsin dataset with only 120 labeled nodes can generalize to 30 new graphs with an average accuracy of 67.26%, surpassing not only all inductive baselines, but also strong transductive methods trained separately on each of the 30 test graphs.

Floorplanning for systems-on-a-chip (SoCs) and its sub-systems is a crucial and non-trivial step of the physical design flow. It represents a difficult combinatorial optimization problem. A typical large scale SoC with 120 partitions generates a search-space of ~ 10250. As novel machine learning (ML) approaches emerge to tackle such problems, there is a growing need for a modern benchmark that comprises a large training dataset and performance metrics that better reflect real-world constraints and objectives compared to existing benchmarks. To address this need, we present FloorSet - two comprehensive datasets of synthetic fixed-outline floorplan layouts that reflect the distribution of real SoCs. Each dataset has 1M training samples and 100 test samples where each sample is a synthetic floor-plan. FloorSet-Prime comprises fully-abutted rectilinear partitions and near-optimal wire-length. A simplified dataset that reflects early design phases, FloorSet-Lite comprises rectangular partitions, with < 5% white-space and near-optimal wire-length. Both datasets define hard constraints seen in modern design flows such as shape constraints, edge-affinity, grouping constraints, and pre-placement constraints. FloorSet is intended to spur fundamental research on large-scale constrained optimization problems. Crucially, FloorSet alleviates the core issue of reproducibility in modern ML driven solutions to such problems. FloorSet is available as an open-source repository for the research community1.

Foundation models in language and vision have the ability to run inference on any textual and visual inputs thanks to the transferable representations such as a vocabulary of tokens in language. Knowledge graphs (KGs) have different entity and relation vocabularies that generally do not overlap. The key challenge of designing foundation models on KGs is to learn such transferable representations that enable inference on any graph with arbitrary entity and relation vocabularies. In this work, we make a step towards such foundation models and present ULTRA, an approach for learning universal and transferable graph representations. ULTRA builds relational representations as a function conditioned on their interactions. Such a conditioning strategy allows a pre-trained ULTRA model to inductively generalize to any unseen KG with any relation vocabulary and to be fine-tuned on any graph. Conducting link prediction experiments on 57 different KGs, we find that the zero-shot inductive inference performance of a single pre-trained ULTRA model on unseen graphs of various sizes is often on par or better than strong baselines trained on specific graphs. Fine-tuning further boosts the performance.

Class imbalance is a well-recognized challenge in GNN-based fraud detection. Traditional methods like re-sampling and re-weighting address this issue by balancing class distribution. However, node class balancing with simple re-sampling or re-weighting may greatly distort the data distributions and eventually lead to the ineffective performance of GNNs. In this paper, we propose a novel approach named Graph Generative Node Augmentation (GGA), which improves GNN-based fraud detection models by augmenting synthetic nodes of the minority class. GGA utilizes the GAN framework to synthesize node features and related edges of fake fraudulent nodes. To introduce greater variety in the generated nodes, we employ an MLP for feature generation. We also introduce an attention module to encode feature-level information before graph convolutional layers for edge generation. Our empirical results on two real-world fraud datasets demonstrate that GGA improves the performance of GNN-based fraud detection models by a large margin with much fewer nodes than traditional class balance methods, and outperforms recent graph augmentation methods with the same number of synthetic nodes.

Dynamic scene graph generation from a video is challenging due to the temporal dynamics of the scene and the inherent temporal fluctuations of predictions. We hypothesize that capturing long-term temporal dependencies is the key to effective generation of dynamic scene graphs. We propose to learn the long-term dependencies in a video by capturing the object-level consistency and inter-object relationship dynamics over object-level long-term tracklets using transformers. Experimental results demonstrate that our Dynamic Scene Graph Detection Transformer (DSG- DETR) outperforms state-of-the-art methods by a significant margin on the benchmark dataset Action Genome. Our ablation studies validate the effectiveness of each component of the proposed approach. The source code is available at https://github.com/Shengyu-Feng/DS G-DETR.

Recent improvements in the performance of state-of-the-art (SOTA) methods for Graph Representational Learning (GRL) have come at the cost of significant computational resource requirements for training, e.g., for calculating gradients via backprop over many data epochs. Meanwhile, Singular Value Decomposition (SVD) can find closed-form solutions to convex problems, using merely a handful of epochs. In this paper, we make GRL more computationally tractable for those with modest hardware. We design a framework that computes SVD of *implicitly* defined matrices, and apply this framework to several GRL tasks. For each task, we derive first-order approximation of a SOTA model, where we design (expensive-to-store) matrix $\mathbf{M}$ and train the model, in closed-form, via SVD of $\mathbf{M}$, without calculating entries of $\mathbf{M}$. By converging to a unique point in one step, and without calculating gradients, our models show competitive empirical test performance over various graphs such as article citation and biological interaction networks. More importantly, SVD can initialize a deeper model, that is architected to be non-linear almost everywhere, though behaves linearly when its parameters reside on a hyperplane, onto which SVD initializes. The deeper model can then be fine-tuned within only a few epochs. Overall, our algorithm trains hundreds of times faster than state-of-the-art methods, while competing on test empirical performance. We open-source our implementation at: https://github.com/samihaija/isvd

Modern deep neural networks are typically highly overparameterized. Pruning techniques are able to remove a significant fraction of network parameters with little loss in accuracy. Recently, techniques based on dynamic reallocation of non-zero parameters have emerged, allowing direct training of sparse networks without having to pre-train a large dense model. Here we present a novel dynamic sparse reparameterization method that addresses the limitations of previous techniques such as high computational cost and the need for manual configuration of the number of free parameters allocated to each layer. We evaluate the performance of dynamic reallocation methods in training deep convolutional networks and show that our method outperforms previous static and dynamic reparameterization methods, yielding the best accuracy for a fixed parameter budget, on par with accuracies obtained by iteratively pruning a pre-trained dense model. We further investigated the mechanisms underlying the superior generalization performance of the resultant sparse networks. We found that neither the structure, nor the initialization of the non-zero parameters were sufficient to explain the superior performance. Rather, effective learning crucially depended on the continuous exploration of the sparse network structure space during training. Our work suggests that exploring structural degrees of freedom during training is more effective than adding extra parameters to the network.

Convolutional neural networks (CNNs) have become the dominant neural network architecture for solving many state-of-the-art (SOA) visual processing tasks. Even though graphical processing units are most often used in training and deploying CNNs, their power efficiency is less than 10 GOp/s/W for single-frame runtime inference. We propose a flexible and efficient CNN accelerator architecture called NullHop that implements SOA CNNs useful for low-power and low-latency application scenarios. NullHop exploits the sparsity of neuron activations in CNNs to accelerate the computation and reduce memory requirements. The flexible architecture allows high utilization of available computing resources across kernel sizes ranging from 1×1 to 7×7. NullHop can process up to 128 input and 128 output feature maps per layer in a single pass. We implemented the proposed architecture on a Xilinx Zynq field-programmable gate array (FPGA) platform and presented the results showing how our implementation reduces external memory transfers and compute time in five different CNNs ranging from small ones up to the widely known large VGG16 and VGG19 CNNs. Postsynthesis simulations using Mentor Modelsim in a 28-nm process with a clock frequency of 500 MHz show that the VGG19 network achieves over 450 GOp/s. By exploiting sparsity, NullHop achieves an efficiency of 368%, maintains over 98% utilization of the multiply-accumulate units, and achieves a power efficiency of over 3 TOp/s/W in a core area of 6.3 mm2. As further proof of NullHop's usability, we interfaced its FPGA implementation with a neuromorphic event camera for real-time interactive demonstrations.

Many recent generative models make use of neural networks to transform the probability distribution of a simple low-dimensional noise process into the complex distribution of the data. This raises the question of whether biological networks operate along similar principles to implement a probabilistic model of the environment through transformations of intrinsic noise processes. The intrinsic neural and synaptic noise processes in biological networks, however, are quite different from the noise processes used in current abstract generative networks. This, together with the discrete nature of spikes and local circuit interactions among the neurons, raises several difficulties when using recent generative modeling frameworks to train biologically motivated models. In this letter, we show that a biologically motivated model based on multilayer winner-take-all circuits and stochastic synapses admits an approximate analytical description. This allows us to use the proposed networks in a variational learning setting where stochastic backpropagation is used to optimize a lower bound on the data log likelihood, thereby learning a generative model of the data. We illustrate the generality of the proposed networks and learning technique by using them in a structured output prediction task and a semisupervised learning task. Our results extend the domain of application of modern stochastic network architectures to networks where synaptic transmission failure is the principal noise mechanism.

Gradient descent training techniques are remarkably successful in training analog-valued artificial neural networks (ANNs). Such training techniques, however, do not transfer easily to spiking networks due to the spike generation hard nonlinearity and the discrete nature of spike communication. We show that in a feedforward spiking network that uses a temporal coding scheme where information is encoded in spike times instead of spike rates, the network input-output relation is differentiable almost everywhere. Moreover, this relation is piecewise linear after a transformation of variables. Methods for training ANNs thus carry directly to the training of such spiking networks as we show when training on the permutation invariant MNIST task. In contrast to rate-based spiking networks that are often used to approximate the behavior of ANNs, the networks we present spike much more sparsely and their behavior cannot be directly approximated by conventional ANNs. Our results highlight a new approach for controlling the behavior of spiking networks with realistic temporal dynamics, opening up the potential for using these networks to process spike patterns with complex temporal information.

Memristors have emerged as promising, area-efficient, nano-scale devices for implementing models of synaptic plasticity in hybrid CMOS-memristor neuromorphic architectures. These architectures aim at reproducing the learning capabilities of biological networks by emulating the complex dynamics of biological neurons and synapses. However, to maximize the density of these elements in crossbar arrays, learning circuits have often been limited to the implementation of simple spike timing-dependent plasticity (STDP) mechanisms. We propose novel hybrid CMOS-memristor circuits that reproduce more effective and realistic plasticity rules which depend on the timing of the pre-synaptic input spike and on the state of the post-synaptic neuron, and which allow the integration of dense crossbar memristor arrays. To implement these plasticity rules in memristor crossbar arrays, the circuits driving the memristors' post-synaptic terminals actively sense the activity on the pre-synaptic terminals to apply the appropriate stimulation waveforms across the memristors. We illustrate the advantages of this scheme by using it to implement a spike-based perceptron plasticity r ule.

Gamma-band rhythmic inhibition is a ubiquitous phenomenon in neural circuits, yet its computational role remains elusive. We show that a model of gamma-band rhythmic inhibition allows networks of coupled cortical circuit motifs to search for network configurations that best reconcile external inputs with an internal consistency model encoded in the network connectivity. We show that Hebbian plasticity allows the networks to learn the consistency model by example. The search dynamics driven by rhythmic inhibition enable the described networks to solve difficult constraint satisfaction problems without making assumptions about the form of stochastic fluctuations in the network. We show that the search dynamics are well approximated by a stochastic sampling process. We use the described networks to reproduce perceptual multistability phenomena with switching times that are a good match to experimental data and show that they provide a general neural framework that can be used to model other perceptual inference phenomena.

To endow large scale VLSI networks of spiking neurons with learning abilities it is important to develop compact and low power circuits that implement synaptic plasticity mechanisms. In this paper we present an analog/digital Spike-Timing Dependent Plasticity (STDP) circuit that changes its internal state in a continuous analog way on short biologically plausible time scales and drives its weight to one of two possible bi-stable states on long time scales. We highlight the differences and improvements over previously proposed circuits and demonstrate the performance of the new circuit using data measured from a chip fabricated using a standard 180nm CMOS process. Finally we discuss the use of stochastic learning methods that can best exploit the properties of this circuit for implementing robust machine-learning algorithms.

Understanding the sequence generation and learning mechanisms used by recurrent neural networks in the nervous system is an important problem that has been studied extensively. However, most of the models proposed in the literature are either not compatible with neuroanatomy and neurophysiology experimental findings, or are not robust to noise and rely on fine tuning of the parameters. In this work, we propose a novel model of sequence learning and generation that is based on the interactions among multiple asymmetrically coupled winner-take-all (WTA) circuits. The network architecture is consistent with mammalian cortical connectivity data and uses realistic neuronal and synaptic dynamics that give rise to noise-robust patterns of sequential activity. The novel aspect of the network we propose lies in its ability to produce robust patterns of sequential activity that can be halted, resumed, and readily modulated by external input, and in its ability to make use of realistic plastic synapses to learn and reproduce the arbitrary input-imposed sequential patterns. Sequential activity takes the form of a single activity bump that stably propagates through multiple WTA circuits along one of a number of possible paths. Because the network can be configured to either generate spontaneous sequences or wait for external inputs to trigger a transition in the sequence, it provides the basis for creating state-dependent perception-action loops. We first analyze a rate-based approximation of the proposed spiking network to highlight the relevant features of the network dynamics and then show numerical simulation results with spiking neurons, realistic conductance-based synapses, and spike-timing dependent plasticity (STDP) rules to validate the rate-based model.

We present a recurrent neuronal network, modeled as a continuous-time dynamical system, that can solve constraint satisfaction problems. Discrete variables are represented by coupled Winner-Take-All (WTA) networks, and their values are encoded in localized patterns of oscillations that are learned by the recurrent weights in these networks. Constraints over the variables are encoded in the network connectivity. Although there are no sources of noise, the network can escape from local optima in its search for solutions that satisfy all constraints by modifying the effective network connectivity through oscillations. If there is no solution that satisfies all constraints, the network state changes in a pseudo-random manner and its trajectory approximates a sampling procedure that selects a variable assignment with a probability that increases with the fraction of constraints satisfied by this assignment. External evidence, or input to the network, can force variables to specific values. When new inputs are applied, the network re-evaluates the entire set of variables in its search for the states that satisfy the maximum number of constraints, while being consistent with the external input. Our results demonstrate that the proposed network architecture can perform a deterministic search for the optimal solution to problems with non-convex cost functions. The network is inspired by canonical microcircuit models of the cortex and suggests possible dynamical mechanisms to solve constraint satisfaction problems that can be present in biological networks, or implemented in neuromorphic electronic circuits.