Marc'Aurelio Ranzato

Yutian Chen, Xingyou Song, Chansoo Lee et al. · deepmind

Meta-learning hyperparameter optimization (HPO) algorithms from prior experiments is a promising approach to improve optimization efficiency over objective functions from a similar distribution. However, existing methods are restricted to learning from experiments sharing the same set of hyperparameters. In this paper, we introduce the OptFormer, the first text-based Transformer HPO framework that provides a universal end-to-end interface for jointly learning policy and function prediction when trained on vast tuning data from the wild, such as Google's Vizier database, one of the world's largest HPO datasets. Our extensive experiments demonstrate that the OptFormer can simultaneously imitate at least 7 different HPO algorithms, which can be further improved via its function uncertainty estimates. Compared to a Gaussian Process, the OptFormer also learns a robust prior distribution for hyperparameter response functions, and can thereby provide more accurate and better calibrated predictions. This work paves the path to future extensions for training a Transformer-based model as a general HPO optimizer.

Jorg Bornschein, Alexandre Galashov, Ross Hemsley et al. · deepmind

A shared goal of several machine learning communities like continual learning, meta-learning and transfer learning, is to design algorithms and models that efficiently and robustly adapt to unseen tasks. An even more ambitious goal is to build models that never stop adapting, and that become increasingly more efficient through time by suitably transferring the accrued knowledge. Beyond the study of the actual learning algorithm and model architecture, there are several hurdles towards our quest to build such models, such as the choice of learning protocol, metric of success and data needed to validate research hypotheses. In this work, we introduce the Never-Ending VIsual-classification Stream (NEVIS'22), a benchmark consisting of a stream of over 100 visual classification tasks, sorted chronologically and extracted from papers sampled uniformly from computer vision proceedings spanning the last three decades. The resulting stream reflects what the research community thought was meaningful at any point in time, and it serves as an ideal test bed to assess how well models can adapt to new tasks, and do so better and more efficiently as time goes by. Despite being limited to classification, the resulting stream has a rich diversity of tasks from OCR, to texture analysis, scene recognition, and so forth. The diversity is also reflected in the wide range of dataset sizes, spanning over four orders of magnitude. Overall, NEVIS'22 poses an unprecedented challenge for current sequential learning approaches due to the scale and diversity of tasks, yet with a low entry barrier as it is limited to a single modality and well understood supervised learning problems. Moreover, we provide a reference implementation including strong baselines and an evaluation protocol to compare methods in terms of their trade-off between accuracy and compute.

Massimo Caccia, Alexandre Galashov, Arthur Douillard et al. · deepmind

The field of transfer learning is undergoing a significant shift with the introduction of large pretrained models which have demonstrated strong adaptability to a variety of downstream tasks. However, the high computational and memory requirements to finetune or use these models can be a hindrance to their widespread use. In this study, we present a solution to this issue by proposing a simple yet effective way to trade computational efficiency for asymptotic performance which we define as the performance a learning algorithm achieves as compute tends to infinity. Specifically, we argue that zero-shot structured pruning of pretrained models allows them to increase compute efficiency with minimal reduction in performance. We evaluate our method on the Nevis'22 continual learning benchmark that offers a diverse set of transfer scenarios. Our results show that pruning convolutional filters of pretrained models can lead to more than 20% performance improvement in low computational regimes.

Adam Fisch, Amal Rannen-Triki, Razvan Pascanu et al. · deepmind

As the application space of language models continues to evolve, a natural question to ask is how we can quickly adapt models to new tasks. We approach this classic question from a continual learning perspective, in which we aim to continue fine-tuning models trained on past tasks on new tasks, with the goal of "transferring" relevant knowledge. However, this strategy also runs the risk of doing more harm than good, i.e., negative transfer. In this paper, we construct a new benchmark of task sequences that target different possible transfer scenarios one might face, such as a sequence of tasks with high potential of positive transfer, high potential for negative transfer, no expected effect, or a mixture of each. An ideal learner should be able to maximally exploit information from all tasks that have any potential for positive transfer, while also avoiding the negative effects of any distracting tasks that may confuse it. We then propose a simple, yet effective, learner that satisfies many of our desiderata simply by leveraging a selective strategy for initializing new models from past task checkpoints. Still, limitations remain, and we hope this benchmark can help the community to further build and analyze such learners.

Arthur Douillard, Qixuan Feng, Andrei A. Rusu et al.

Large language models (LLM) have become a critical component in many applications of machine learning. However, standard approaches to training LLM require a large number of tightly interconnected accelerators, with devices exchanging gradients and other intermediate states at each optimization step. While it is difficult to build and maintain a single computing cluster hosting many accelerators, it might be easier to find several computing clusters each hosting a smaller number of devices. In this work, we propose a distributed optimization algorithm, Distributed Low-Communication (DiLoCo), that enables training of language models on islands of devices that are poorly connected. The approach is a variant of federated averaging, where the number of inner steps is large, the inner optimizer is AdamW, and the outer optimizer is Nesterov momentum. On the widely used C4 dataset, we show that DiLoCo on 8 workers performs as well as fully synchronous optimization while communicating 500 times less. DiLoCo exhibits great robustness to the data distribution of each worker. It is also robust to resources becoming unavailable over time, and vice versa, it can seamlessly leverage resources that become available during training.

Arthur Douillard, Keith Rush, Yani Donchev et al.

Modern large-scale language model pre-training relies heavily on the single program multiple data (SPMD) paradigm, which requires tight coupling across accelerators. Due to this coupling, transient slowdowns, hardware failures, and synchronization overhead stall the entire computation, wasting significant compute time at scale. While recent distributed methods like DiLoCo reduced communication bandwidth, they remained fundamentally synchronous and vulnerable to these system stalls. To address this, we introduce Decoupled DiLoCo, an evolution of the DiLoCo framework designed to break the lock-step synchronization barrier and go beyond SPMD to maximize training goodput. Decoupled DiLoCo partitions compute across multiple independent ``learners'' that execute local inner optimization steps. These learners asynchronously communicate parameter fragments to a central synchronizer, which circumvents failed or straggling learners by aggregating updates using a minimum quorum, an adaptive grace window, and dynamic token-weighted merging. Inspired by ``chaos engineering'', we achieve significantly improved training efficiency in failure-prone environments with millions of simulated chips with strictly zero global downtime, while maintaining competitive model performance across text and vision tasks, for both dense and mixture-of-expert architectures.

Lucio M. Dery, Abram L. Friesen, Nando De Freitas et al.

As machine learning permeates more industries and models become more expensive and time consuming to train, the need for efficient automated hyperparameter optimization (HPO) has never been more pressing. Multi-step planning based approaches to hyperparameter optimization promise improved efficiency over myopic alternatives by more effectively balancing out exploration and exploitation. However, the potential of these approaches has not been fully realized due to their technical complexity and computational intensity. In this work, we leverage recent advances in Transformer-based, natural-language-interfaced hyperparameter optimization to circumvent these barriers. We build on top of the recently proposed OptFormer which casts both hyperparameter suggestion and target function approximation as autoregressive generation thus making planning via rollouts simple and efficient. We conduct extensive exploration of different strategies for performing multi-step planning on top of the OptFormer model to highlight its potential for use in constructing non-myopic HPO strategies.

Zeyu Huang, Adhiguna Kuncoro, Qixuan Feng et al.

Most existing large language models (LLMs) are expensive to adapt after deployment, especially when a task requires newly produced information or niche domain knowledge. Recent work has shown that, by manipulating and optimizing their context, LLMs can be tailored to downstream tasks without updating their weights. However, most existing methods remain closed-loop, relying solely on the model's intrinsic knowledge. In this paper, we equip these context optimizers with Wikipedia search and browser tools for active information seeking. We show that naively adding these tools to a standard sequential context optimization pipeline can actually degrade performance compared to baselines. However, when paired with a search-based training procedure that maintains and prunes multiple candidate contexts, active information seeking delivers consistent and substantial gains. We demonstrate these improvements across diverse domains, including low-resource translation (Flores+), health scenarios (HealthBench), and reasoning-heavy tasks (LiveCodeBench and Humanity's Last Exam). Furthermore, our method proves to be data-efficient, robust across different hyperparameters, and capable of generating effective textual contexts that generalize well across different models.

Lucas Caccia, Jing Xu, Myle Ott et al.

In many practical applications of machine learning data arrives sequentially over time in large chunks. Practitioners have then to decide how to allocate their computational budget in order to obtain the best performance at any point in time. Online learning theory for convex optimization suggests that the best strategy is to use data as soon as it arrives. However, this might not be the best strategy when using deep non-linear networks, particularly when these perform multiple passes over each chunk of data rendering the overall distribution non i.i.d.. In this paper, we formalize this learning setting in the simplest scenario in which each data chunk is drawn from the same underlying distribution, and make a first attempt at empirically answering the following questions: How long should the learner wait before training on the newly arrived chunks? What architecture should the learner adopt? Should the learner increase capacity over time as more data is observed? We probe this learning setting using convolutional neural networks trained on classic computer vision benchmarks as well as a large transformer model trained on a large-scale language modeling task. Code is available at \url{www.github.com/facebookresearch/ALMA}.

Tianxiao Shen, Myle Ott, Michael Auli et al.

Mixture models trained via EM are among the simplest, most widely used and well understood latent variable models in the machine learning literature. Surprisingly, these models have been hardly explored in text generation applications such as machine translation. In principle, they provide a latent variable to control generation and produce a diverse set of hypotheses. In practice, however, mixture models are prone to degeneracies---often only one component gets trained or the latent variable is simply ignored. We find that disabling dropout noise in responsibility computation is critical to successful training. In addition, the design choices of parameterization, prior distribution, hard versus soft EM and online versus offline assignment can dramatically affect model performance. We develop an evaluation protocol to assess both quality and diversity of generations against multiple references, and provide an extensive empirical study of several mixture model variants. Our analysis shows that certain types of mixture models are more robust and offer the best trade-off between translation quality and diversity compared to variational models and diverse decoding approaches.\footnote{Code to reproduce the results in this paper is available at \url{https://github.com/pytorch/fairseq}}

Francisco Guzmán, Peng-Jen Chen, Myle Ott et al.

For machine translation, a vast majority of language pairs in the world are considered low-resource because they have little parallel data available. Besides the technical challenges of learning with limited supervision, it is difficult to evaluate methods trained on low-resource language pairs because of the lack of freely and publicly available benchmarks. In this work, we introduce the FLoRes evaluation datasets for Nepali-English and Sinhala-English, based on sentences translated from Wikipedia. Compared to English, these are languages with very different morphology and syntax, for which little out-of-domain parallel data is available and for which relatively large amounts of monolingual data are freely available. We describe our process to collect and cross-check the quality of translations, and we report baseline performance using several learning settings: fully supervised, weakly supervised, semi-supervised, and fully unsupervised. Our experiments demonstrate that current state-of-the-art methods perform rather poorly on this benchmark, posing a challenge to the research community working on low-resource MT. Data and code to reproduce our experiments are available at https://github.com/facebookresearch/flores.

Bo Liu, Rachita Chhaparia, Arthur Douillard et al.

Local stochastic gradient descent (Local-SGD), also referred to as federated averaging, is an approach to distributed optimization where each device performs more than one SGD update per communication. This work presents an empirical study of {\it asynchronous} Local-SGD for training language models; that is, each worker updates the global parameters as soon as it has finished its SGD steps. We conduct a comprehensive investigation by examining how worker hardware heterogeneity, model size, number of workers, and optimizer could impact the learning performance. We find that with naive implementations, asynchronous Local-SGD takes more iterations to converge than its synchronous counterpart despite updating the (global) model parameters more frequently. We identify momentum acceleration on the global parameters when worker gradients are stale as a key challenge. We propose a novel method that utilizes a delayed Nesterov momentum update and adjusts the workers' local training steps based on their computation speed. This approach, evaluated with models up to 150M parameters on the C4 dataset, matches the performance of synchronous Local-SGD in terms of perplexity per update step, and significantly surpasses it in terms of wall clock time.

Arthur Douillard, Yanislav Donchev, Keith Rush et al.

Training of large language models (LLMs) is typically distributed across a large number of accelerators to reduce training time. Since internal states and parameter gradients need to be exchanged at each and every single gradient step, all devices need to be co-located using low-latency high-bandwidth communication links to support the required high volume of exchanged bits. Recently, distributed algorithms like DiLoCo have relaxed such co-location constraint: accelerators can be grouped into ``workers'', where synchronizations between workers only occur infrequently. This in turn means that workers can afford being connected by lower bandwidth communication links without affecting learning quality. However, in these methods, communication across workers still requires the same peak bandwidth as before, as the synchronizations require all parameters to be exchanged across all workers. In this paper, we improve DiLoCo in three ways. First, we synchronize only subsets of parameters in sequence, rather than all at once, which greatly reduces peak bandwidth. Second, we allow workers to continue training while synchronizing, which decreases wall clock time. Third, we quantize the data exchanged by workers, which further reduces bandwidth across workers. By properly combining these modifications, we show experimentally that we can distribute training of billion-scale parameters and reach similar quality as before, but reducing required bandwidth by two orders of magnitude.

Arthur Douillard, Qixuan Feng, Andrei A. Rusu et al.

Progress in machine learning (ML) has been fueled by scaling neural network models. This scaling has been enabled by ever more heroic feats of engineering, necessary for accommodating ML approaches that require high bandwidth communication between devices working in parallel. In this work, we propose a co-designed modular architecture and training approach for ML models, dubbed DIstributed PAth COmposition (DiPaCo). During training, DiPaCo distributes computation by paths through a set of shared modules. Together with a Local-SGD inspired optimization (DiLoCo) that keeps modules in sync with drastically reduced communication, Our approach facilitates training across poorly connected and heterogeneous workers, with a design that ensures robustness to worker failures and preemptions. At inference time, only a single path needs to be executed for each input, without the need for any model compression. We consider this approach as a first prototype towards a new paradigm of large-scale learning, one that is less synchronous and more modular. Our experiments on the widely used C4 benchmark show that, for the same amount of training steps but less wall-clock time, DiPaCo exceeds the performance of a 1 billion-parameter dense transformer language model by choosing one of 256 possible paths, each with a size of 150 million parameters.

Naman Goyal, Cynthia Gao, Vishrav Chaudhary et al.

One of the biggest challenges hindering progress in low-resource and multilingual machine translation is the lack of good evaluation benchmarks. Current evaluation benchmarks either lack good coverage of low-resource languages, consider only restricted domains, or are low quality because they are constructed using semi-automatic procedures. In this work, we introduce the FLORES-101 evaluation benchmark, consisting of 3001 sentences extracted from English Wikipedia and covering a variety of different topics and domains. These sentences have been translated in 101 languages by professional translators through a carefully controlled process. The resulting dataset enables better assessment of model quality on the long tail of low-resource languages, including the evaluation of many-to-many multilingual translation systems, as all translations are multilingually aligned. By publicly releasing such a high-quality and high-coverage dataset, we hope to foster progress in the machine translation community and beyond.

Tom Veniat, Ludovic Denoyer, Marc'Aurelio Ranzato

Existing literature in Continual Learning (CL) has focused on overcoming catastrophic forgetting, the inability of the learner to recall how to perform tasks observed in the past. There are however other desirable properties of a CL system, such as the ability to transfer knowledge from previous tasks and to scale memory and compute sub-linearly with the number of tasks. Since most current benchmarks focus only on forgetting using short streams of tasks, we first propose a new suite of benchmarks to probe CL algorithms across these new axes. Finally, we introduce a new modular architecture, whose modules represent atomic skills that can be composed to perform a certain task. Learning a task reduces to figuring out which past modules to re-use, and which new modules to instantiate to solve the current task. Our learning algorithm leverages a task-driven prior over the exponential search space of all possible ways to combine modules, enabling efficient learning on long streams of tasks. Our experiments show that this modular architecture and learning algorithm perform competitively on widely used CL benchmarks while yielding superior performance on the more challenging benchmarks we introduce in this work.

Lajanugen Logeswaran, Ann Lee, Myle Ott et al.

Few-shot algorithms aim at learning new tasks provided only a handful of training examples. In this work we investigate few-shot learning in the setting where the data points are sequences of tokens and propose an efficient learning algorithm based on Transformers. In the simplest setting, we append a token to an input sequence which represents the particular task to be undertaken, and show that the embedding of this token can be optimized on the fly given few labeled examples. Our approach does not require complicated changes to the model architecture such as adapter layers nor computing second order derivatives as is currently popular in the meta-learning and few-shot learning literature. We demonstrate our approach on a variety of tasks, and analyze the generalization properties of several model variants and baseline approaches. In particular, we show that compositional task descriptors can improve performance. Experiments show that our approach works at least as well as other methods, while being more computationally efficient.

Sandeep Subramanian, Ronan Collobert, Marc'Aurelio Ranzato et al.

We investigate multi-scale transformer language models that learn representations of text at multiple scales, and present three different architectures that have an inductive bias to handle the hierarchical nature of language. Experiments on large-scale language modeling benchmarks empirically demonstrate favorable likelihood vs memory footprint trade-offs, e.g. we show that it is possible to train a hierarchical variant with 30 layers that has 23% smaller memory footprint and better perplexity, compared to a vanilla transformer with less than half the number of layers, on the Toronto BookCorpus. We analyze the advantages of learned representations at multiple scales in terms of memory footprint, compute time, and perplexity, which are particularly appealing given the quadratic scaling of transformers' run time and memory usage with respect to sequence length.

Yuntian Deng, Anton Bakhtin, Myle Ott et al.

Text generation is ubiquitous in many NLP tasks, from summarization, to dialogue and machine translation. The dominant parametric approach is based on locally normalized models which predict one word at a time. While these work remarkably well, they are plagued by exposure bias due to the greedy nature of the generation process. In this work, we investigate un-normalized energy-based models (EBMs) which operate not at the token but at the sequence level. In order to make training tractable, we first work in the residual of a pretrained locally normalized language model and second we train using noise contrastive estimation. Furthermore, since the EBM works at the sequence level, we can leverage pretrained bi-directional contextual representations, such as BERT and RoBERTa. Our experiments on two large language modeling datasets show that residual EBMs yield lower perplexity compared to locally normalized baselines. Moreover, generation via importance sampling is very efficient and of higher quality than the baseline models according to human evaluation.

Anton Bakhtin, Yuntian Deng, Sam Gross et al.

Current large-scale auto-regressive language models display impressive fluency and can generate convincing text. In this work we start by asking the question: Can the generations of these models be reliably distinguished from real text by statistical discriminators? We find experimentally that the answer is affirmative when we have access to the training data for the model, and guardedly affirmative even if we do not. This suggests that the auto-regressive models can be improved by incorporating the (globally normalized) discriminators into the generative process. We give a formalism for this using the Energy-Based Model framework, and show that it indeed improves the results of the generative models, measured both in terms of perplexity and in terms of human evaluation.

Peng-Jen Chen, Jiajun Shen, Matt Le et al.

This paper describes Facebook AI's submission to the WAT 2019 Myanmar-English translation task. Our baseline systems are BPE-based transformer models. We explore methods to leverage monolingual data to improve generalization, including self-training, back-translation and their combination. We further improve results by using noisy channel re-ranking and ensembling. We demonstrate that these techniques can significantly improve not only a system trained with additional monolingual data, but even the baseline system trained exclusively on the provided small parallel dataset. Our system ranks first in both directions according to human evaluation and BLEU, with a gain of over 8 BLEU points above the second best system.

Junxian He, Jiatao Gu, Jiajun Shen et al.

Self-training is one of the earliest and simplest semi-supervised methods. The key idea is to augment the original labeled dataset with unlabeled data paired with the model's prediction (i.e. the pseudo-parallel data). While self-training has been extensively studied on classification problems, in complex sequence generation tasks (e.g. machine translation) it is still unclear how self-training works due to the compositionality of the target space. In this work, we first empirically show that self-training is able to decently improve the supervised baseline on neural sequence generation tasks. Through careful examination of the performance gains, we find that the perturbation on the hidden states (i.e. dropout) is critical for self-training to benefit from the pseudo-parallel data, which acts as a regularizer and forces the model to yield close predictions for similar unlabeled inputs. Such effect helps the model correct some incorrect predictions on unlabeled data. To further encourage this mechanism, we propose to inject noise to the input space, resulting in a "noisy" version of self-training. Empirical study on standard machine translation and text summarization benchmarks shows that noisy self-training is able to effectively utilize unlabeled data and improve the performance of the supervised baseline by a large margin.

Jiajun Shen, Peng-Jen Chen, Matt Le et al.

While we live in an increasingly interconnected world, different places still exhibit strikingly different cultures and many events we experience in our every day life pertain only to the specific place we live in. As a result, people often talk about different things in different parts of the world. In this work we study the effect of local context in machine translation and postulate that particularly in low resource settings this causes the domains of the source and target language to greatly mismatch, as the two languages are often spoken in further apart regions of the world with more distinctive cultural traits and unrelated local events. We first formalize the concept of source-target domain mismatch, propose a metric to quantify it, and provide empirical evidence corroborating our intuition that organic text produced by people speaking very different languages exhibits the most dramatic differences. We conclude with an empirical study of how source-target domain mismatch affects training of machine translation systems for low resource language pairs. In particular, we find that it severely affects back-translation, but the degradation can be alleviated by combining back-translation with self-training and by increasing the relative amount of target side monolingual data.

Sergey Edunov, Myle Ott, Marc'Aurelio Ranzato et al.

Back-translation is a widely used data augmentation technique which leverages target monolingual data. However, its effectiveness has been challenged since automatic metrics such as BLEU only show significant improvements for test examples where the source itself is a translation, or translationese. This is believed to be due to translationese inputs better matching the back-translated training data. In this work, we show that this conjecture is not empirically supported and that back-translation improves translation quality of both naturally occurring text as well as translationese according to professional human translators. We provide empirical evidence to support the view that back-translation is preferred by humans because it produces more fluent outputs. BLEU cannot capture human preferences because references are translationese when source sentences are natural text. We recommend complementing BLEU with a language model score to measure fluency.

Guillaume Lample, Alexandre Sablayrolles, Marc'Aurelio Ranzato et al.

This paper introduces a structured memory which can be easily integrated into a neural network. The memory is very large by design and significantly increases the capacity of the architecture, by up to a billion parameters with a negligible computational overhead. Its design and access pattern is based on product keys, which enable fast and exact nearest neighbor search. The ability to increase the number of parameters while keeping the same computational budget lets the overall system strike a better trade-off between prediction accuracy and computation efficiency both at training and test time. This memory layer allows us to tackle very large scale language modeling tasks. In our experiments we consider a dataset with up to 30 billion words, and we plug our memory layer in a state-of-the-art transformer-based architecture. In particular, we found that a memory augmented model with only 12 layers outperforms a baseline transformer model with 24 layers, while being twice faster at inference time. We release our code for reproducibility purposes.

Anton Bakhtin, Sam Gross, Myle Ott et al.

Energy-based models (EBMs), a.k.a. un-normalized models, have had recent successes in continuous spaces. However, they have not been successfully applied to model text sequences. While decreasing the energy at training samples is straightforward, mining (negative) samples where the energy should be increased is difficult. In part, this is because standard gradient-based methods are not readily applicable when the input is high-dimensional and discrete. Here, we side-step this issue by generating negatives using pre-trained auto-regressive language models. The EBM then works in the residual of the language model; and is trained to discriminate real text from text generated by the auto-regressive models. We investigate the generalization ability of residual EBMs, a pre-requisite for using them in other applications. We extensively analyze generalization for the task of classifying whether an input is machine or human generated, a natural task given the training loss and how we mine negatives. Overall, we observe that EBMs can generalize remarkably well to changes in the architecture of the generators producing negatives. However, EBMs exhibit more sensitivity to the training set used by such generators.

Senthil Purushwalkam, Maximilian Nickel, Abhinav Gupta et al.

One of the hallmarks of human intelligence is the ability to compose learned knowledge into novel concepts which can be recognized without a single training example. In contrast, current state-of-the-art methods require hundreds of training examples for each possible category to build reliable and accurate classifiers. To alleviate this striking difference in efficiency, we propose a task-driven modular architecture for compositional reasoning and sample efficient learning. Our architecture consists of a set of neural network modules, which are small fully connected layers operating in semantic concept space. These modules are configured through a gating function conditioned on the task to produce features representing the compatibility between the input image and the concept under consideration. This enables us to express tasks as a combination of sub-tasks and to generalize to unseen categories by reweighting a set of small modules. Furthermore, the network can be trained efficiently as it is fully differentiable and its modules operate on small sub-spaces. We focus our study on the problem of compositional zero-shot classification of object-attribute categories. We show in our experiments that current evaluation metrics are flawed as they only consider unseen object-attribute pairs. When extending the evaluation to the generalized setting which accounts also for pairs seen during training, we discover that naive baseline methods perform similarly or better than current approaches. However, our modular network is able to outperform all existing approaches on two widely-used benchmark datasets.

Arslan Chaudhry, Marcus Rohrbach, Mohamed Elhoseiny et al.

In continual learning (CL), an agent learns from a stream of tasks leveraging prior experience to transfer knowledge to future tasks. It is an ideal framework to decrease the amount of supervision in the existing learning algorithms. But for a successful knowledge transfer, the learner needs to remember how to perform previous tasks. One way to endow the learner the ability to perform tasks seen in the past is to store a small memory, dubbed episodic memory, that stores few examples from previous tasks and then to replay these examples when training for future tasks. In this work, we empirically analyze the effectiveness of a very small episodic memory in a CL setup where each training example is only seen once. Surprisingly, across four rather different supervised learning benchmarks adapted to CL, a very simple baseline, that jointly trains on both examples from the current task as well as examples stored in the episodic memory, significantly outperforms specifically designed CL approaches with and without episodic memory. Interestingly, we find that repetitive training on even tiny memories of past tasks does not harm generalization, on the contrary, it improves it, with gains between 7\% and 17\% when the memory is populated with a single example per class.

Arslan Chaudhry, Marc'Aurelio Ranzato, Marcus Rohrbach et al.

In lifelong learning, the learner is presented with a sequence of tasks, incrementally building a data-driven prior which may be leveraged to speed up learning of a new task. In this work, we investigate the efficiency of current lifelong approaches, in terms of sample complexity, computational and memory cost. Towards this end, we first introduce a new and a more realistic evaluation protocol, whereby learners observe each example only once and hyper-parameter selection is done on a small and disjoint set of tasks, which is not used for the actual learning experience and evaluation. Second, we introduce a new metric measuring how quickly a learner acquires a new skill. Third, we propose an improved version of GEM (Lopez-Paz & Ranzato, 2017), dubbed Averaged GEM (A-GEM), which enjoys the same or even better performance as GEM, while being almost as computationally and memory efficient as EWC (Kirkpatrick et al., 2016) and other regularization-based methods. Finally, we show that all algorithms including A-GEM can learn even more quickly if they are provided with task descriptors specifying the classification tasks under consideration. Our experiments on several standard lifelong learning benchmarks demonstrate that A-GEM has the best trade-off between accuracy and efficiency.

Sandeep Subramanian, Guillaume Lample, Eric Michael Smith et al.

The dominant approach to unsupervised "style transfer" in text is based on the idea of learning a latent representation, which is independent of the attributes specifying its "style". In this paper, we show that this condition is not necessary and is not always met in practice, even with domain adversarial training that explicitly aims at learning such disentangled representations. We thus propose a new model that controls several factors of variation in textual data where this condition on disentanglement is replaced with a simpler mechanism based on back-translation. Our method allows control over multiple attributes, like gender, sentiment, product type, etc., and a more fine-grained control on the trade-off between content preservation and change of style with a pooling operator in the latent space. Our experiments demonstrate that the fully entangled model produces better generations, even when tested on new and more challenging benchmarks comprising reviews with multiple sentences and multiple attributes.

Guillaume Lample, Myle Ott, Alexis Conneau et al.

Machine translation systems achieve near human-level performance on some languages, yet their effectiveness strongly relies on the availability of large amounts of parallel sentences, which hinders their applicability to the majority of language pairs. This work investigates how to learn to translate when having access to only large monolingual corpora in each language. We propose two model variants, a neural and a phrase-based model. Both versions leverage a careful initialization of the parameters, the denoising effect of language models and automatic generation of parallel data by iterative back-translation. These models are significantly better than methods from the literature, while being simpler and having fewer hyper-parameters. On the widely used WMT'14 English-French and WMT'16 German-English benchmarks, our models respectively obtain 28.1 and 25.2 BLEU points without using a single parallel sentence, outperforming the state of the art by more than 11 BLEU points. On low-resource languages like English-Urdu and English-Romanian, our methods achieve even better results than semi-supervised and supervised approaches leveraging the paucity of available bitexts. Our code for NMT and PBSMT is publicly available.

Anton Bakhtin, Arthur Szlam, Marc'Aurelio Ranzato et al.

It is often the case that the best performing language model is an ensemble of a neural language model with n-grams. In this work, we propose a method to improve how these two models are combined. By using a small network which predicts the mixture weight between the two models, we adapt their relative importance at each time step. Because the gating network is small, it trains quickly on small amounts of held out data, and does not add overhead at scoring time. Our experiments carried out on the One Billion Word benchmark show a significant improvement over the state of the art ensemble without retraining of the basic modules.

Myle Ott, Michael Auli, David Grangier et al.

Machine translation is a popular test bed for research in neural sequence-to-sequence models but despite much recent research, there is still a lack of understanding of these models. Practitioners report performance degradation with large beams, the under-estimation of rare words and a lack of diversity in the final translations. Our study relates some of these issues to the inherent uncertainty of the task, due to the existence of multiple valid translations for a single source sentence, and to the extrinsic uncertainty caused by noisy training data. We propose tools and metrics to assess how uncertainty in the data is captured by the model distribution and how it affects search strategies that generate translations. Our results show that search works remarkably well but that models tend to spread too much probability mass over the hypothesis space. Next, we propose tools to assess model calibration and show how to easily fix some shortcomings of current models. As part of this study, we release multiple human reference translations for two popular benchmarks.

Sergey Edunov, Myle Ott, Michael Auli et al.

There has been much recent work on training neural attention models at the sequence-level using either reinforcement learning-style methods or by optimizing the beam. In this paper, we survey a range of classical objective functions that have been widely used to train linear models for structured prediction and apply them to neural sequence to sequence models. Our experiments show that these losses can perform surprisingly well by slightly outperforming beam search optimization in a like for like setup. We also report new state of the art results on both IWSLT'14 German-English translation as well as Gigaword abstractive summarization. On the larger WMT'14 English-French translation task, sequence-level training achieves 41.5 BLEU which is on par with the state of the art.

Guillaume Lample, Alexis Conneau, Ludovic Denoyer et al.

Machine translation has recently achieved impressive performance thanks to recent advances in deep learning and the availability of large-scale parallel corpora. There have been numerous attempts to extend these successes to low-resource language pairs, yet requiring tens of thousands of parallel sentences. In this work, we take this research direction to the extreme and investigate whether it is possible to learn to translate even without any parallel data. We propose a model that takes sentences from monolingual corpora in two different languages and maps them into the same latent space. By learning to reconstruct in both languages from this shared feature space, the model effectively learns to translate without using any labeled data. We demonstrate our model on two widely used datasets and two language pairs, reporting BLEU scores of 32.8 and 15.1 on the Multi30k and WMT English-French datasets, without using even a single parallel sentence at training time.

Alexis Conneau, Guillaume Lample, Marc'Aurelio Ranzato et al.

State-of-the-art methods for learning cross-lingual word embeddings have relied on bilingual dictionaries or parallel corpora. Recent studies showed that the need for parallel data supervision can be alleviated with character-level information. While these methods showed encouraging results, they are not on par with their supervised counterparts and are limited to pairs of languages sharing a common alphabet. In this work, we show that we can build a bilingual dictionary between two languages without using any parallel corpora, by aligning monolingual word embedding spaces in an unsupervised way. Without using any character information, our model even outperforms existing supervised methods on cross-lingual tasks for some language pairs. Our experiments demonstrate that our method works very well also for distant language pairs, like English-Russian or English-Chinese. We finally describe experiments on the English-Esperanto low-resource language pair, on which there only exists a limited amount of parallel data, to show the potential impact of our method in fully unsupervised machine translation. Our code, embeddings and dictionaries are publicly available.

David Lopez-Paz, Marc'Aurelio Ranzato

One major obstacle towards AI is the poor ability of models to solve new problems quicker, and without forgetting previously acquired knowledge. To better understand this issue, we study the problem of continual learning, where the model observes, once and one by one, examples concerning a sequence of tasks. First, we propose a set of metrics to evaluate models learning over a continuum of data. These metrics characterize models not only by their test accuracy, but also in terms of their ability to transfer knowledge across tasks. Second, we propose a model for continual learning, called Gradient Episodic Memory (GEM) that alleviates forgetting, while allowing beneficial transfer of knowledge to previous tasks. Our experiments on variants of the MNIST and CIFAR-100 datasets demonstrate the strong performance of GEM when compared to the state-of-the-art.

Guillaume Lample, Neil Zeghidour, Nicolas Usunier et al.

This paper introduces a new encoder-decoder architecture that is trained to reconstruct images by disentangling the salient information of the image and the values of attributes directly in the latent space. As a result, after training, our model can generate different realistic versions of an input image by varying the attribute values. By using continuous attribute values, we can choose how much a specific attribute is perceivable in the generated image. This property could allow for applications where users can modify an image using sliding knobs, like faders on a mixing console, to change the facial expression of a portrait, or to update the color of some objects. Compared to the state-of-the-art which mostly relies on training adversarial networks in pixel space by altering attribute values at train time, our approach results in much simpler training schemes and nicely scales to multiple attributes. We present evidence that our model can significantly change the perceived value of the attributes while preserving the naturalness of images.

Sam Gross, Marc'Aurelio Ranzato, Arthur Szlam

Training convolutional networks (CNN's) that fit on a single GPU with minibatch stochastic gradient descent has become effective in practice. However, there is still no effective method for training large CNN's that do not fit in the memory of a few GPU cards, or for parallelizing CNN training. In this work we show that a simple hard mixture of experts model can be efficiently trained to good effect on large scale hashtag (multilabel) prediction tasks. Mixture of experts models are not new (Jacobs et. al. 1991, Collobert et. al. 2003), but in the past, researchers have had to devise sophisticated methods to deal with data fragmentation. We show empirically that modern weakly supervised data sets are large enough to support naive partitioning schemes where each data point is assigned to a single expert. Because the experts are independent, training them in parallel is easy, and evaluation is cheap for the size of the model. Furthermore, we show that we can use a single decoding layer for all the experts, allowing a unified feature embedding space. We demonstrate that it is feasible (and in fact relatively painless) to train far larger models than could be practically trained with standard CNN architectures, and that the extra capacity can be well used on current datasets.

Sam Wiseman, Sumit Chopra, Marc'Aurelio Ranzato et al.

While Truncated Back-Propagation through Time (BPTT) is the most popular approach to training Recurrent Neural Networks (RNNs), it suffers from being inherently sequential (making parallelization difficult) and from truncating gradient flow between distant time-steps. We investigate whether Target Propagation (TPROP) style approaches can address these shortcomings. Unfortunately, extensive experiments suggest that TPROP generally underperforms BPTT, and we end with an analysis of this phenomenon, and suggestions for future work.

Joost van Amersfoort, Anitha Kannan, Marc'Aurelio Ranzato et al.

In this work we propose a simple unsupervised approach for next frame prediction in video. Instead of directly predicting the pixels in a frame given past frames, we predict the transformations needed for generating the next frame in a sequence, given the transformations of the past frames. This leads to sharper results, while using a smaller prediction model. In order to enable a fair comparison between different video frame prediction models, we also propose a new evaluation protocol. We use generated frames as input to a classifier trained with ground truth sequences. This criterion guarantees that models scoring high are those producing sequences which preserve discriminative features, as opposed to merely penalizing any deviation, plausible or not, from the ground truth. Our proposed approach compares favourably against more sophisticated ones on the UCF-101 data set, while also being more efficient in terms of the number of parameters and computational cost.

Jiwei Li, Alexander H. Miller, Sumit Chopra et al.

A good dialogue agent should have the ability to interact with users by both responding to questions and by asking questions, and importantly to learn from both types of interaction. In this work, we explore this direction by designing a simulator and a set of synthetic tasks in the movie domain that allow such interactions between a learner and a teacher. We investigate how a learner can benefit from asking questions in both offline and online reinforcement learning settings, and demonstrate that the learner improves when asking questions. Finally, real experiments with Mechanical Turk validate the approach. Our work represents a first step in developing such end-to-end learned interactive dialogue agents.

Jiwei Li, Alexander H. Miller, Sumit Chopra et al.

An important aspect of developing conversational agents is to give a bot the ability to improve through communicating with humans and to learn from the mistakes that it makes. Most research has focused on learning from fixed training sets of labeled data rather than interacting with a dialogue partner in an online fashion. In this paper we explore this direction in a reinforcement learning setting where the bot improves its question-answering ability from feedback a teacher gives following its generated responses. We build a simulator that tests various aspects of such learning in a synthetic environment, and introduce models that work in this regime. Finally, real experiments with Mechanical Turk validate the approach.

Marc'Aurelio Ranzato, Sumit Chopra, Michael Auli et al.

Many natural language processing applications use language models to generate text. These models are typically trained to predict the next word in a sequence, given the previous words and some context such as an image. However, at test time the model is expected to generate the entire sequence from scratch. This discrepancy makes generation brittle, as errors may accumulate along the way. We address this issue by proposing a novel sequence level training algorithm that directly optimizes the metric used at test time, such as BLEU or ROUGE. On three different tasks, our approach outperforms several strong baselines for greedy generation. The method is also competitive when these baselines employ beam search, while being several times faster.

Mark Tygert, Arthur Szlam, Soumith Chintala et al.

The conventional classification schemes -- notably multinomial logistic regression -- used in conjunction with convolutional networks (convnets) are classical in statistics, designed without consideration for the usual coupling with convnets, stochastic gradient descent, and backpropagation. In the specific application to supervised learning for convnets, a simple scale-invariant classification stage turns out to be more robust than multinomial logistic regression, appears to result in slightly lower errors on several standard test sets, has similar computational costs, and features precise control over the actual rate of learning. "Scale-invariant" means that multiplying the input values by any nonzero scalar leaves the output unchanged.

Tomas Mikolov, Armand Joulin, Sumit Chopra et al.

Recurrent neural network is a powerful model that learns temporal patterns in sequential data. For a long time, it was believed that recurrent networks are difficult to train using simple optimizers, such as stochastic gradient descent, due to the so-called vanishing gradient problem. In this paper, we show that learning longer term patterns in real data, such as in natural language, is perfectly possible using gradient descent. This is achieved by using a slight structural modification of the simple recurrent neural network architecture. We encourage some of the hidden units to change their state slowly by making part of the recurrent weight matrix close to identity, thus forming kind of a longer term memory. We evaluate our model in language modeling experiments, where we obtain similar performance to the much more complex Long Short Term Memory (LSTM) networks (Hochreiter & Schmidhuber, 1997).

Grégoire Mesnil, Tomas Mikolov, Marc'Aurelio Ranzato et al.

Sentiment analysis is a common task in natural language processing that aims to detect polarity of a text document (typically a consumer review). In the simplest settings, we discriminate only between positive and negative sentiment, turning the task into a standard binary classification problem. We compare several ma- chine learning approaches to this problem, and combine them to achieve the best possible results. We show how to use for this task the standard generative lan- guage models, which are slightly complementary to the state of the art techniques. We achieve strong results on a well-known dataset of IMDB movie reviews. Our results are easily reproducible, as we publish also the code needed to repeat the experiments. This should simplify further advance of the state of the art, as other researchers can combine their techniques with ours with little effort.

Yaniv Taigman, Ming Yang, Marc'Aurelio Ranzato et al.

Scaling machine learning methods to very large datasets has attracted considerable attention in recent years, thanks to easy access to ubiquitous sensing and data from the web. We study face recognition and show that three distinct properties have surprising effects on the transferability of deep convolutional networks (CNN): (1) The bottleneck of the network serves as an important transfer learning regularizer, and (2) in contrast to the common wisdom, performance saturation may exist in CNN's (as the number of training samples grows); we propose a solution for alleviating this by replacing the naive random subsampling of the training set with a bootstrapping process. Moreover, (3) we find a link between the representation norm and the ability to discriminate in a target domain, which sheds lights on how such networks represent faces. Based on these discoveries, we are able to improve face recognition accuracy on the widely used LFW benchmark, both in the verification (1:1) and identification (1:N) protocols, and directly compare, for the first time, with the state of the art Commercially-Off-The-Shelf system and show a sizable leap in performance.

Marc'Aurelio Ranzato

Current automatic vision systems face two major challenges: scalability and extreme variability of appearance. First, the computational time required to process an image typically scales linearly with the number of pixels in the image, therefore limiting the resolution of input images to thumbnail size. Second, variability in appearance and pose of the objects constitute a major hurdle for robust recognition and detection. In this work, we propose a model that makes baby steps towards addressing these challenges. We describe a learning based method that recognizes objects through a series of glimpses. This system performs an amount of computation that scales with the complexity of the input rather than its number of pixels. Moreover, the proposed method is potentially more robust to changes in appearance since its parameters are learned in a data driven manner. Preliminary experiments on a handwritten dataset of digits demonstrate the computational advantages of this approach.

Omry Yadan, Keith Adams, Yaniv Taigman et al.

In this work we evaluate different approaches to parallelize computation of convolutional neural networks across several GPUs.