cheng

cheng, 1 month ago to random

I've finally started reading "The Bezzle" by the wonderful @pluralistic. Did we really speak like that 20 years ago? "How's life in the punctuation factory?", “Looks like they’re in full swing. Good. I hate being the first ones at a party.”

I really enjoyed the story so far (still reading), with a very straightforward identification of bad guys by saying that "There is no crime in X".

In this age of algorithmic recommendations, I support Cory Doctorow's crusade to do network marketing!

ayo, 1 month ago to random

Let’s promote alternative search engines! Comment your favorite if you have one :)

cheng, 3 months ago to llm

This is an excellent argument by Jennifer Listgarten about why Large Language Models #LLM like #chatGPT are not a silver bullet for scientific discovery. I am also motivated to study #ExperimentalDesign using #MachineLearning for the reasons Jennifer argues in this paper.

We need better data in #science.

https://www.nature.com/articles/s41587-023-02103-0

skinnylatte, 4 months ago (edited 4 months ago) to random

Every time we have Malay food for dinner in San Francisco I think about how we wouldn’t be able to exist here without the pastes we cart back when we visit. They’re expensive and they have to be kept chill or frozen, but.. it also tastes like a nasi Padang store back home, which is such a luxury.

When we have time we make stuff from scratch but sometimes we don’t have all the things we need (bunga kantan, daun kunyit..)

https://www.wahyubrand.com/collections/paste

#Food #TootSea

cheng, 5 months ago to machinelearning

Chris Bishop has a new #MachineLearning book!

Bishop and Bishop have published a new book on the concepts behind Deep Learning.
https://www.bishopbook.com/

I've taught from "Pattern Recognition and Machine Learning" for almost a decade, and I really like Chris Bishop's writing. I haven't yet read this new book, and would welcome thoughts from people who take a look.

cheng, 7 months ago to machinelearning

A nice survey paper (led by @bendfulcher) about ways to measure similarity/distance between a pair of multivariate #TimeSeries. It explains an organising structure around the different bits of literature that compares time series.

#MachineLearning
https://www.nature.com/articles/s43588-023-00519-x

cheng, 8 months ago to machinelearning

We propose a new family of probability densities that have closed form normalising constants. Our densities use two layer neural networks as parameters, and strictly generalise exponential families. We show that the squared norm can be integrated in closed form, resulting in the normalizing constant. We call the densities Squared Neural Family (#SNEFY), which are closed under conditioning.

Accepted at #NeurIPS2023. #MachineLearning #Bayesian #GaussianProcess

https://arxiv.org/abs/2305.13552

cheng, 9 months ago to random

#ICML2023 outstanding papers award (6 of them):
https://icml.cc/Conferences/2023/Awards

• theoretical analysis of non-smooth stochastic convex optimization
• watermarking output of large language models
• learning Boolean functions, generalising to unseen ones
• Follow the Regularized Leader (FTRL) variations for optimising imperfect information games
• Markov Chain Monte Carlo for graphs - via self-repellent random walks
• Adaptive experimental design - a new objective called algorithmic information ratio

cheng, 10 months ago to ArtificialIntelligence

I'm fascinated by the discussion about how #ArtificialIntelligence will take our jobs. Susskind and Susskind have a book called "The Future of Professions", this is a short version. While professions look very different, they all solve the same problem: nobody can know everything. As we move from a print based society, to an internet society, how do we organise work? It is useful to think about the tasks that change.

https://www.amphilsoc.org/sites/default/files/2018-11/attachments/Susskind%20and%20Susskind.pdf

#MachineLearning #DigitalDisruption #FutureOfWork

cheng, 11 months ago to machinelearning

Consider a Transformer as a kernel machine. Look at the extension of kernels to an asymmetric form. Motivated by empirical observation of low rank properties of self-attention, find a link between kernel Singular Value Decomposition and self attention.

#MachineLearning #KernelMethods #Attention #Transformers

https://arxiv.org/abs/2305.19798

msteenhagen, 11 months ago to random

Just had furu for the first time. (It’s a fermented tofu, stored in a jar with flavoured brine.)

This is really the most remarkable flavour I’ve tasted in years—in a good way.

It’s close to blue cheese, and goes extremely well on a cracker or stirred through a dish. It is rather pungent on its own, like a very old Stilton x2.

cheng, 1 year ago to random

Theoretical analysis of #MachineLearning decision making. #Fairness viewed as #SocialWelfare should also take into account the alternative decision (called #counterfactual outcomes). This paper proposes a framework which models two kinds of actors: (1) a "principal" who is the central organisation; (2) a set of "agents" who provides the service under consideration. The social welfare goal is to design reward functions with high utility for the principal.

https://arxiv.org/abs/2305.14595

cheng, 1 year ago to random

The international #conference on #MachineLearning ( #icml2023 ) has 30 associated workshops, and all of them are calling for contributions right now.

https://icml.cc/virtual/2023/events/workshop

cheng, 1 year ago to random

Content moderation and #RecommenderSystems are the most widely deployed #MachineLearning based action taking systems. By action taking, I mean an algorithm that suggests a particular decision that impacts human actions. Using this view of recommender systems, we can consider how to increase the #diversity of online content by re-ranking recommended items to encourage creation of diverse content in the long term.

https://arxiv.org/abs/2302.04336