Andrej Karpathy, Justin Johnson, Li Fei-Fei
2015 · arXiv
Recurrent Neural Networks (RNNs), and specifically a variant with Long Short-Term Memory (LSTM), are enjoying renewed interest as a result of successful applications in a wide range of machine learning problems that involve sequential data. However, while LSTMs provide exceptional results in practice, the source of their performance and their limitations remain rather poorly understood. Using character-level language models as an interpretable testbed, we aim to bridge this gap by providing an analysis of their representations, predictions and error types. In particular, our experiments reveal the existence of interpretable cells that keep track of long-range dependencies such as line lengths, quotes and brackets. Moreover, our comparative analysis with finite horizon n-gram models traces the source of the LSTM improvements to long-range structural dependencies. Finally, we provide analysis of the remaining errors and suggests areas for further study.
Metadata: arXiv. Abstracts retain their original authors’ rights.
Original HN comments, with labels from simple text rules. A label describes the whole comment and may miss context or irony. Read the discussion to make up your own mind. About the labels
4 selected excerpts · 4 total mentions
I tried to find the where I heard that Radford was inspired by that blog post, but the closest thing I found is that in the "Sentiment Neuron" paper (Learning to Generate Reviews and Discovering Sentiment: https://arxiv.org/pdf/1704.01444.pdf), in the "Discussion and Future Work" section they mention this Karpathy paper from 2015: Visualizing and Understanding Recurrent Networks https://arxiv.org/abs/1506.02078
Yikes, you are right… I just linked a private repo. '^^ I have poked the rest of the group and it seems that at least a tweet was made [1] – but not much else remains. Describing it from memory, we ran ELMo and BERT on Wikipedia and then allowed similarity search between a query and showed heat maps to a matched context. Nothing particularly deep compared to yours that go into the transformer “machinery”, but I think it captures very well how most Question Answering models still operate: Embed query and contexts in a high-dimensional space, compare, find semantically plausible span, and done! [1]: https://twitter.com/Johannes_Welbl/status/106530965474036121... Work and…
articles like yours has truly had an impact on me, even though they are largely qualitative. We always say “Turing complete” this and “Turing complete” that, but theoretical statements such as this have little practical utility to me as we all know that what can be learnt and what is learnt are two very different things. For example, “Visualizing and Understanding Recurrent Networks” by Karpathy et al. (2015) [2] that you list as inspiration blew my mind in terms of for example neurons that monotonically decrease from the sentence start. I remember Karpathy giving a talk on it in London and what struck me was how he simply had gone to manually inspect the neurons manually (heresy!) as there were only a few thousand of them any way. That playfulness, truly admirable. [2]: https://arxiv.org/abs/1506.02078 Another anecdote, now from “Attention Is All You Need” by Vaswani et al. (2017) [3] where I was far from sold on Transformers as a model until Uszkoreit gave a talk at an invitation-only summit where he showed those cherry-picked attention heads that “flipped” based on whether an object was animate or not. I approached him after the talk and asked why it was not in the paper as it was awesome! Maybe I am biased because I give a large role to intuition in science, but analysis such as this is far more valuable to me as a researcher than yet another point of BLEU or a 10th dataset. Again, my bias, but I feel that there is a need for new ways of thinking in terms of both “hard” empiricism and “soft” analysis in machine learning as we seemingly are now having to mature given the attention we are receiving. [3]: https://arxiv.org/abs/1706.03762 Apologies if I am rambling, it is midnight now and I barely slept last night.
This one is a character level language model. The Microsoft one is per token. This one should be the closest to the Microsoft model: https://miltos.allamanis.com/publicationfiles/allamanis2018l... Btw, simple neural networks are far from doing the best for code. The kind of accuracy LSTM gives is already achieved with good n-gram models. E.g. see this one on a Linux code prediction: https://arxiv.org/pdf/1506.02078.pdf Also, there are other techniques that improve on top of the n-gram models: https://files.sri.inf.ethz.ch/website/papers/charmodel-iclr2...
So, this is definitely good work, and I don't have any suggestions on how to do it better, but I'm ultimately not sure how useful I find it. I have no intuition for what the histograms for `t` should look like, and so seeing the different histograms for ծ vs `not ծ` doesn't tell me much except that they're different (which is trivially true!). Does anyone else find this sort of visualization useful? Maybe I'm just misunderstanding this. I would love to develop more of an intuition for neuron activations in deep nets- I want to better understand how they work, as right now, the only thing that I do is feed in inputs and look at the outputs, which is wildly inadequate. I…
have the same criticism when it comes to the CNN visualizations that Karpathy et. al [1] came up with. They're cool, but I don't find them that useful. I hope I don't sound too critical. I'm really glad that people are doing this sort of work, as I think that it's incredibly important to the advancement of deep learning, and I think that this work is well-done, but I don't find it personally useful. [1]: https://arxiv.org/abs/1506.02078
More from the same topics.
Sébastien Bubeck, Varun Chandrasekaran, Ronen Eldan +11
Artificial intelligence (AI) researchers have been developing and refining large language models (LLMs) that exhibit remarkable capabilities across a variety of domains and tasks, challenging our understanding of learning and cognition.
Tom B. Brown, Benjamin Mann, Nick Ryder +28
Recent work has demonstrated substantial gains on many NLP tasks and benchmarks by pre-training on a large corpus of text followed by fine-tuning on a specific task.
Chiyuan Zhang, Samy Bengio, Moritz Hardt +2
Despite their massive size, successful deep artificial neural networks can exhibit a remarkably small difference between training and test performance.
Mariusz Bojarski, Davide Del Testa, Daniel Dworakowski +10
We trained a convolutional neural network (CNN) to map raw pixels from a single front-facing camera directly to steering commands.