Upon noticing a reminder email about upcoming reviews due for a conference 3 days ago, I realized that I haven’t received any prior notifications about the assignments and thus haven’t started reading the papers yet. It is not a good form to miss reviews, especially for a conference that I previously chaired, so I spent an intense long weekend reviewing these submissions.
2025-05-05
2025-03-22
Trend following II
Joining a hot research trend is like buying a stock when the price is high.
These neural network pioneers (like Hinton, LeCun, etc.) started their works not when the field was hot, but when it was considered a dead end. And they stuck with it for decades until the right circumstances (e.g., GPU compute finally became feasible). Now, every time a new idea (about generative X or whatever) pops in the head, there might already be 10 arXiv papers on it.
It is important to know any significant new advancements, but it is even more important to figure out what and how to apply these into our own research and be able to formulate our own (unique) vision.
2025-02-18
Holiday time for exploratory fun
Monday the 17th is a holiday and I was pondering whether I should start a work task that I aimed to finish by the end of the week or spend time on more exploratory research. I decided to go for the latter as I realized that during my usual workdays I often feel enough obligation towards my collaborators for the former and I am the only one who can push myself for the latter.
My decision turned out to be correct for this particular one, as on Monday I had some fun tinkering with \href{https://glicol.org/}{computer music language} (that I have zero background with) and on Tuesday I immediately hit on issues that required responses from my colleagues upon starting the work task.
And I also spent some time doing some silly doodling:
2025-01-26
Single-authored SIGGRAPH submission from a recent collaborator whom I have yet to meet in person
I just looked at a single-authored SIGGRAPH submission from a recent collaborator (with a first-authored SIGGRAPH paper before even getting into a graduate degree program), intentionally after the deadline to ensure that I wouldn’t be tempted to contribute anything to his solo effort (as I just did) before the review process.
I like the work and hope it will get accepted so that we could finally meet in person at the conference.
(Hopefully the Canadian visa would be easier to obtain than the US equivalence, as otherwise he would not have any co-author to present the paper.)
2025-01-16
The devil sheds Prada
I used to concentrate my paper submissions to a venue that gained prestige by accepting only the most novel ideas, but its prestige has been diluted with more incremental works over the years to the point that it has higher acceptance rate than alternatives (in an overlapping but different field) that I now submit to more.
2024-12-14
Art papers
I took a look at the art papers in SIGGRAPH Asia 2024 primarily due to the unusual preface and, to a lesser extent, the much lower (and thus more precious) number (28 only!) and acceptance rate (15.9%) than the technical papers program.
Unlike the technical papers which have clearer review criteria (novelty, importance, utility, readability, reproducibility, ethics, etc.), I agree with the art papers chair that it is not clear what is an art paper, not least a good one. I found some of the papers interesting in the technical or philosophical perspectives, while some others left me scratching my head. But I guess this is what art is in general – more about expressing/sharing novel viewpoints/experiences than scientific discoveries or practical utilities.
2024-11-16
Most important skills for ML R&D
Secure compute resources and figure out the compatible combinations of Python package versions, and the rest will follow naturally. 😈
2024-09-22
Unexpected intellectual discovery
There are several aspects of scientific research that excite me: solving a challenging problem, publishing a paper, shipping a product, creating a demo/artwork, presenting a talk, reviewing a submission, serving on a committee, attending a conference, meeting people with shared interests, working with collaborators.
But what excites me the most is discovering new ideas and insights, especially those that are surprising or unexpected.
Looking back at my research career, here are some examples:
- My PhD thesis work on texture synthesis – fixed neighborhoods, over manifold surfaces, from multiple sources (blending and 3D volumes from 2D views), and order-independent/random-accessible synthesis.
- Tile-based texture mapping, in particular the math and algorithms for packing and random access.
- Parallel white noise and blue noise generation, analogous to the order-independent texture synthesis.
- Inverse texture synthesis, basically a reversal of traditional (forward) texture synthesis.
- Nonlinear Beam Tracing on a GPU \cite{Liu:2011:NBT} – it is possible to use rasterization to achieve some ray tracing effects.
- Nonlinear revision control for images – revision control can be applied beyond code and text!
- Differential domain analysis for non-uniform sampling – the trigonometric expansion of the power spectrum depends only on the relative sample positions, and the subsequent synthesis method of point sampling with general noise spectrum.
- Improving light field camera sample design with irregularity and aberration – randomness/noise helps with computational photography.
- Autocomplete painting repetitions, hand-drawn animations, and 3D sculpting – the use of workflows for analysis and prediction.
- Mapping virtual and physical reality and the subsequent work on leveraging temporary blindness to avoid warping at all for redirected walking \cite{Sun:2018:TVR}.
In retrospect, I tend to have more focus, and likely as a consequence, more innovation during the earlier stage of my career, and more breadth (in terms of topics and applications) towards the later stage.
But it is really the unexpected intellectual discovery that keeps me excited, which I missed.
2024-07-09
Bad reviews during a rebuttal are like bad calls during a game
By “bad” reviews, I meant those that are short or unclear/uninformative, which is different/orthogonal from “negative” reviews which criticize or reject the work.
If a review is bad because it is too short, it wouldn’t bother me at all. I can write a correspondingly short portion of my rebuttal addressing that review, or even ignore it altogether.
If a review is bad because it is unclear, I can ask for clarification in the rebuttal, and make a few educated guesses about the exact questions or criticisms followed by my answers.
The last thing I would do is to complain about the bad reviews in the rebuttal, or (worse) contact the paper chairs/committee. These are like complaining about bad calls during a game. What would you expect the paper chairs/committee to do? Ask for an additional review or the bad reviewer to submit a better review? That would take time, and even if you receive an updated review, it might be too late for you to address it before the end of the rebuttal period.