Publications (22)
Finite Versus Infinite Neural Networks: an Empirical Study
Jaehoon Lee, Samuel S. Schoenholz, Jeffrey Pennington +4
We perform a careful, thorough, and large scale empirical study of the correspondence between wide neural networks and kernel methods. By doing so, we resolve a variety of open que…
Exact posterior distributions of wide Bayesian neural networks
Jiri Hron, Yasaman Bahri, Roman Novak +2
Recent work has shown that the prior over functions induced by a deep Bayesian neural network (BNN) behaves as a Gaussian process (GP) as the width of all layers becomes large. How…
Scaling Exponents Across Parameterizations and Optimizers
Katie Everett, Lechao Xiao, Mitchell Wortsman +8
Robust and effective scaling of models from small to large width typically requires the precise adjustment of many algorithmic and architectural details, such as parameterization a…
Deep Neural Networks as Gaussian Processes
Jaehoon Lee, Yasaman Bahri, Roman Novak +3
It has long been known that a single-layer fully-connected neural network with an i.i.d. prior over its parameters is equivalent to a Gaussian process (GP), in the limit of infinit…
Sensitivity and Generalization in Neural Networks: an Empirical Study
Roman Novak, Yasaman Bahri, Daniel A. Abolafia +2
In practice it is often found that large over-parameterized neural networks generalize better than their smaller counterparts, an observation that appears to conflict with classica…
Fast Neural Kernel Embeddings for General Activations
Insu Han, Amir Zandieh, Jaehoon Lee +3
Infinite width limit has shed light on generalization and optimization aspects of deep learning by establishing connections between neural networks and kernel methods. Despite thei…
Bayesian Deep Convolutional Networks with Many Channels are Gaussian Processes
Roman Novak, Lechao Xiao, Jaehoon Lee +6
There is a previously identified equivalence between wide fully connected neural networks (FCNs) and Gaussian processes (GPs). This equivalence enables, for instance, test set pred…
On the infinite width limit of neural networks with a standard parameterization
Jascha Sohl-Dickstein, Roman Novak, Samuel S. Schoenholz +1
There are currently two parameterizations used to derive fixed kernels corresponding to infinite width neural networks, the NTK (Neural Tangent Kernel) parameterization and the nai…
Exploring the Neural Algorithm of Artistic Style
Yaroslav Nikulin, Roman Novak
We explore the method of style transfer presented in the article "A Neural Algorithm of Artistic Style" by Leon A. Gatys, Alexander S. Ecker and Matthias Bethge (arXiv:1508.06576).…
Beyond the Imitation Game: Quantifying and extrapolating the capabilities of language models
Aarohi Srivastava, Abhinav Rastogi, Abhishek Rao +448
Language models demonstrate both quantitative improvement and new qualitative capabilities with increasing scale. Despite their potentially transformative impact, these new capabil…
Infinite attention: NNGP and NTK for deep attention networks
Jiri Hron, Yasaman Bahri, Jascha Sohl-Dickstein +1
There is a growing amount of literature on the relationship between wide neural networks (NNs) and Gaussian processes (GPs), identifying an equivalence between the two for a variet…
Small-scale proxies for large-scale Transformer training instabilities
Mitchell Wortsman, Peter J. Liu, Lechao Xiao +13
Teams that have trained large Transformer-based models have reported training instabilities at large scale that did not appear when training with the same hyperparameters at smalle…
Wide Neural Networks of Any Depth Evolve as Linear Models Under Gradient Descent
Jaehoon Lee, Lechao Xiao, Samuel S. Schoenholz +4
A longstanding goal in deep learning research has been to precisely characterize training and generalization. However, the often complex loss landscapes of neural networks have mad…
Fast Finite Width Neural Tangent Kernel
Roman Novak, Jascha Sohl-Dickstein, Samuel S. Schoenholz
The Neural Tangent Kernel (NTK), defined as where $\left[\…
Iterative Refinement for Machine Translation
Roman Novak, Michael Auli, David Grangier
Existing machine translation decoding algorithms generate translations in a strictly monotonic fashion and never revisit previous decisions. As a result, earlier mistakes cannot be…
Improving the Neural Algorithm of Artistic Style
Roman Novak, Yaroslav Nikulin
In this work we investigate different avenues of improving the Neural Algorithm of Artistic Style (by Leon A. Gatys, Alexander S. Ecker and Matthias Bethge, arXiv:1508.06576). Whil…
Beyond Human Data: Scaling Self-Training for Problem-Solving with Language Models
Avi Singh, John D. Co-Reyes, Rishabh Agarwal +38
Fine-tuning language models~(LMs) on human-generated data remains a prevalent practice. However, the performance of such models is often limited by the quantity and diversity of hi…
Wide Bayesian neural networks have a simple weight posterior: theory and accelerated sampling
Jiri Hron, Roman Novak, Jeffrey Pennington +1
We introduce repriorisation, a data-dependent reparameterisation which transforms a Bayesian neural network (BNN) posterior to a distribution whose KL divergence to the BNN prior v…
Training Language Models on the Knowledge Graph: Insights on Hallucinations and Their Detectability
Jiri Hron, Laura Culp, Gamaleldin Elsayed +28
While many capabilities of language models (LMs) improve with increased training budget, the influence of scale on hallucinations is not yet fully understood. Hallucinations come i…
Frontier Language Models are not Robust to Adversarial Arithmetic, or "What do I need to say so you agree 2+2=5?
C. Daniel Freeman, Laura Culp, Aaron Parisi +27
We introduce and study the problem of adversarial arithmetic, which provides a simple yet challenging testbed for language model alignment. This problem is comprised of arithmetic…
Dataset Distillation with Infinitely Wide Convolutional Networks
Timothy Nguyen, Roman Novak, Lechao Xiao +1
The effectiveness of machine learning algorithms arises from being able to extract useful features from large amounts of data. As model and dataset sizes increase, dataset distilla…
Neural Tangents: Fast and Easy Infinite Neural Networks in Python
Roman Novak, Lechao Xiao, Jiri Hron +4
Neural Tangents is a library designed to enable research into infinite-width neural networks. It provides a high-level API for specifying complex and hierarchical neural network ar…