When training a machine learning (ML) model, our main aim is usually to get the ‘best’ model out the other end in an unbiased manner. Of course, there are other considerations such as quick training and inference, but mostly we want to be good at predicting the right answer.
A number of factors will affect the quality of our final model, including the chosen architecture, optimiser, and – importantly – the metric we are optimising for. So, how should we pick this metric?
Continue readingTransformers are a very popular architecture for processing sequential data, notably text and (our interest) proteins. Transformers learn more complex patterns with larger models on more data, as demonstrated by models like GPT-4 and ESM-2. Transformers work by updating tokens according to an attention value computed as a weighted sum of all other tokens. In standard implentations this requires computing the product of a query and key matrix which requires O(N2d) computations and, problematically, O(N2) memory for a sequence of length N and an embedding size of d. To speed up Transformers, and to analyze longer sequences, several variants have been proposed which require only O(N) memory. Broadly, these can be divided into sparse methods, softmax-approximators, and memory-efficient Transformers.
Continue readingData can often naturally be represented in a graph format and being able to directly employ a deep learning architecture on that data without finding a different representation is an appealing idea. Graph neural networks (GNNs) have become a standard part of the ML toolbox but navigating the world of different architectures available out-of-the-box can be a daunting task. A great place to start looking for architectures is with PyTorch Geometric, which provides an extensive list of readily available GNN layers and tutorials on how to use them in your standard PyTorch models. There are many things to consider when choosing a GNN layer, but the two considerations that I think are a great place to start are expressiveness and edge feature handling. In general, it is hard to predict what will work best for the task at hand and hence it’s optimal to try a wide range of different layers. This blogpost is meant as a brief introduction for what I would find useful to know before I started using GNNs, and a starting point for exploring the GNN literature.
Continue readingTraining a large transformer model can be a multi-day, if not multi-week, ordeal. Especially if you’re using cloud compute, this can be a very expensive affair, not to mention the environmental impact. It’s therefore worth spending a couple days trying to optimise your training efficiency before embarking on a large scale training run. Here, I’ll run through three strategies you can take which (hopefully) shouldn’t degrade performance, while giving you some free speed. These strategies will also work for any other models using linear layers.
I wont go into too much of the technical detail of any of the techniques, but if you’d like to dig into any of them further I’d highly recommend the Nvidia Deep Learning Performance Guide.
Training with mixed precision can be as simple as adding a few lines of code, depending on your deep learning framework. It also potentially provides the biggest boost to performance of any of these techniques. Training throughput can be increase by up to three-fold with little degradation in performance – and who doesn’t like free speed?
Continue readingTransformers are a very popular architecture in machine learning. While they were first introduced in natural language processing, they have been applied to many fields such as protein folding and design.
Transformers were first introduced in the excellent paper Attention is all you need by Vaswani et al. The paper describes the key elements, including multiheaded attention, and how they come together to create a sequence to sequence model for language translation. The key advance in Attention is all you need is the replacement of all recurrent layers with pure attention + fully connected blocks. Attention is very efficeint to compute and allows for fast comparisons over long distances within a sequence.
One issue, however, is that attention does not natively include a notion of position within a sequence. This means that all tokens could be scrambled and would produce the same result. To overcome this, one can explicitely add a positional encoding to each token. Ideally, such a positional encoding should reflect the relative distance between tokens when computing the query/key comparison such that closer tokens are attended to more than futher tokens. In Attention is all you need, Vaswani et al. propose the slightly mysterious sinusoidal positional encodings which are simply added to the token embeddings:
Last month, a drift of OPIGlets attended the royal society of chemistry’s annual AI in chemistry conference. Co-organised by the group’s very own Garrett Morris and hosted in Churchill College, Cambridge, during a heatwave (!), the two days of conference featured aspects of artificial intelligence and deep machine learning methods to applications in chemistry. The programme included a mixture of keynote talks, panel discussion, oral presentations, flash presentations, posters and opportunities for open debate, networking and discussion amongst participants from academia and industry alike.
Continue readingMultivariate Normal distributions are an essential component of virtually any modern deep learning method—be it to initialise the weights and biases of a neural network, perform variational inference in a probabilistic model, or provide a tractable noise distribution for generative modelling.
What most of us (including—until very recently—me) aren’t aware of, however, is that these Normal distributions begin to look less and less like the characteristic bell curve that we associate them with as their dimensionality increases.
Continue readingWhen designing machine learning algorithms with the aim of accelerating the discovery of novel and more effective therapeutics, we often care deeply about their ability to generalise to new regions of chemical space and accurately predict the properties of molecules that are structurally or functionally dissimilar to the ones we have already explored. To evaluate the performance of algorithms in such an out-of-distribution setting, it is essential that we are able to quantify the data shift that is induced by the train-test splits that we rely on to decide which model to deploy in production.
For our recent ICML 2023 paper Drug Discovery under Covariate Shift with Domain-Informed Prior Distributions over Functions, we chose to quantify the distributional similarity between two sets of molecules through the Maximum Mean Discrepancy (MMD).
Continue readingRecently, I’ve been using a Convolutional Neural Network (CNN), and other methods, to predict the binding affinity of antibodies from their sequence. However, nine months ago, I applied a CNN to a far more important task – distinguishing images of butter from margarine. Please check out the GitHub link below to learn moo-re.
https://github.com/lewis-chinery/AI_cant_believe_its_not_butter
Lucubration – The action or occupation of intensive study originally by candle or lamplight.
Gaslighting – Psychological abuse in which a person or group causes someone to question their own sanity, memories, or perception.
I was recently having a play with Google Bard. Bard, unlike ChatGPT has access to live data. It also undergoes live feedback and quality control. I was hoping to see if it would find me any journals with articles on prion research which I’d previously overlooked.
Me: Please show me some recent articles about prion research.
(Because always be polite to our AI overlords, they’ll remember!)