The conventional lottery ticket hypothesis (LTH) claims that there exists a sparse subnetwork within a dense neural network and a proper random initialization method called the winning ticket, such that it can be trained from scratch to almost as good as the dense counterpart. Meanwhile, the research of LTH in vision transformers (ViTs) is scarcely evaluated. In this paper, we first show that the conventional winning ticket is hard to find at the weight level of ViTs by existing methods. Then, we generalize the LTH for ViTs to input data consisting of image patches inspired by the input dependence of ViTs. That is, there exists a subset of input image patches such that a ViT can be trained from scratch by using only this subset of patches and achieve similar accuracy to the ViTs trained by using all image patches. We call this subset of input patches the em winning tickets, which represent a significant amount of information in the input data. We use a ticket selector to generate the winning tickets based on the informativeness of patches for various types of ViT, including DeiT, LV-ViT, and Swin Transformers. The experiments show that there is a clear difference between the performance of models trained with winning tickets and randomly selected subsets, which verifies our proposed theory. We elaborate on the analogical similarity between our proposed Data-LTH-ViTs and the conventional LTH to further verify the integrity of our theory. The Source codes are available at https://github.com/shawnricecake/vit-lottery-ticket-input.
Analogy-making gives rise to reasoning, abstraction, flexible categorization and counterfactual inference -- abilities lacking in even the best AI systems today. Much research has suggested that analogies are key to non-brittle systems that can adapt to new domains. Despite their importance, analogies received little attention in the NLP community, with most research focusing on simple word analogies. Work that tackled more complex analogies relied heavily on manually constructed, hard-to-scale input representations. In this work, we explore a more realistic, challenging setup: our input is a pair of natural language procedural texts, describing a situation or a process (e.g., how the heart works/how a pump works). Our goal is to automatically extract entities and their relations from the text and find a mapping between the different domains based on relational similarity (e.g., blood is mapped to water). We develop an interpretable, scalable algorithm and demonstrate that it identifies the correct mappings 87% of the time for procedural texts and 94% for stories from cognitive-psychology literature. We show it can extract analogies from a large dataset of procedural texts, achieving 79% precision (analogy prevalence in data: 3%). Lastly, we demonstrate that our algorithm is robust to paraphrasing the input texts.
We investigate the optimal model size and number of tokens for training a transformer language model under a given compute budget. We find that current large language models are significantly undertrained, a consequence of the recent focus on scaling language models whilst keeping the amount of training data constant. By training over \nummodels language models ranging from 70 million to over 16 billion parameters on 5 to 500 billion tokens, we find that for compute-optimal training, the model size and the number of training tokens should be scaled equally: for every doubling of model size the number of training tokens should also be doubled. We test this hypothesis by training a predicted compute-optimal model, \chinchilla, that uses the same compute budget as \gopher but with 70B parameters and 4$\times$ more more data. \chinchilla uniformly and significantly outperforms \Gopher (280B), GPT-3 (175B), Jurassic-1 (178B), and Megatron-Turing NLG (530B) on a large range of downstream evaluation tasks. This also means that \chinchilla uses substantially less compute for fine-tuning and inference, greatly facilitating downstream usage. As a highlight, \chinchilla reaches a state-of-the-art average accuracy of 67.5\% on the MMLU benchmark, greater than a 7\% improvement over \gopher.
Language modelling provides a step towards intelligent communication systems by harnessing large repositories of written human knowledge to better predict and understand the world. In this paper, we present an analysis of Transformer-based language model performance across a wide range of model scales -- from models with tens of millions of parameters up to a 280 billion parameter model called Gopher. These models are evaluated on 152 diverse tasks, achieving state-of-the-art performance across the majority. Gains from scale are largest in areas such as reading comprehension, fact-checking, and the identification of toxic language, but logical and mathematical reasoning see less benefit. We provide a holistic analysis of the training dataset and model's behaviour, covering the intersection of model scale with bias and toxicity. Finally we discuss the application of language models to AI safety and the mitigation of downstream harms.
Much of human learning and inference can be framed within the computational problem of relational generalization. In this project, we propose a Bayesian model that generalizes relational knowledge to novel environments by analogically weighting predictions from previously encountered relational structures. First, we show that this learner outperforms a naive, theory-based learner on relational data derived from random- and Wikipedia-based systems when experience with the environment is small. Next, we show how our formalization of analogical similarity translates to the selection and weighting of analogies. Finally, we combine the analogy- and theory-based learners in a single nonparametric Bayesian model, and show that optimal relational generalization transitions from relying on analogies to building a theory of the novel system with increasing experience in it. Beyond predicting unobserved interactions better than either baseline, this formalization gives a computational-level perspective on the formation and abstraction of analogies themselves.
We introduce a novel evolutionary algorithm (EA) with a semantic network-based representation. For enabling this, we establish new formulations of EA variation operators, crossover and mutation, that we adapt to work on semantic networks. The algorithm employs commonsense reasoning to ensure all operations preserve the meaningfulness of the networks, using ConceptNet and WordNet knowledge bases. The algorithm can be interpreted as a novel memetic algorithm (MA), given that (1) individuals represent pieces of information that undergo evolution, as in the original sense of memetics as it was introduced by Dawkins; and (2) this is different from existing MA, where the word "memetic" has been used as a synonym for local refinement after global optimization. For evaluating the approach, we introduce an analogical similarity-based fitness measure that is computed through structure mapping. This setup enables the open-ended generation of networks analogous to a given base network.
Research in analogical reasoning suggests that higher-order cognitive functions such as abstract reasoning, far transfer, and creativity are founded on recognizing structural similarities among relational systems. Here we integrate theories of analogy with the computational framework of reinforcement learning (RL). We propose a psychology theory that is a computational synergy between analogy and RL, in which analogical comparison provides the RL learning algorithm with a measure of relational similarity, and RL provides feedback signals that can drive analogical learning. Simulation results support the power of this approach.