Comments (6)
Oh wait, I just realized that FFTKDE supports weighting. So we can calculate the mode more accurately by using the square root of the number of reviews.
Note that this time I decided not to sort the values, so instead the order is HRM, HSM, KDE:
[0.3534, 0.3601, 0.5294]
[0.9, 0.9, 0.9759]
[2.3, 2.5861, 2.5091]
[10.9, 10.9, 365.0]
[5.0491, 5.049, 5.0304]
[1.0598, 1.0597, 0.1]
[0.7406, 0.86, 0.8587]
[0.0, 0.0, 0.0]
[1.49, 1.49, 1.4894]
[0.1, 0.1, 0.1]
[0.94, 0.94, 0.9414]
[2.1257, 2.18, 2.1809]
[0.01, 0.01, 0.01]
[0.34, 0.34, 0.3397]
[1.26, 1.26, 2.0]
[0.0, 0.0, 0.0]
[2.61, 2.61, 4.0]
from srs-benchmark.
So here's the final result:
[0.3568, 0.9, 2.5476, 10.9, 5.049, 1.0598, 0.8594, 0.0, 1.49, 0.1, 0.94, 2.1804, 0.01, 0.34, 1.26, 0.0, 2.61]
It's pretty clear that the mode is affected by the default value of a parameter a lot. So I think it's necessary to exclude all values that are equal to their defaults. For example, say S0 is 10.9 for some user. That's the default value. So we need to remove it before calculating the mode.
from srs-benchmark.
Here are the new best modes, after excluding users who have 1 (or more) parameter that is exactly equal to it's default value:
[0.5738, 1.4817, 2.3917, 6.1572, 5.049, 1.0598, 0.7226, 0.0, 1.3714, 0.1, 0.8204, 2.1257, 0.01, 0.2692, 1.1412, 0.0, 2.7572]
from srs-benchmark.
Slightly unrelated. but I noticed that w[12] is distributed pretty much uniformly. All other parameters follow some sort of non-uniform distribution.
from srs-benchmark.
I find that w[15]'s mode is zero. It's a signal that many users press hard when they should press again, because the stability didn't increase (if we didn't clamp it, it could be negative).
from srs-benchmark.
We could allow SInc for "Hard" to be less than 1, in other words, allow stability to decrease, but that will likely lead to a new type of Ease Hell, where the user keeps pressing "Hard" and the intervals get shorter and shorter.
from srs-benchmark.
Related Issues (20)
- Inclusion of any of the boosting models HOT 23
- Remove collecitons of people who misuse Hard from the calculation of the default parameters HOT 4
- collect bad cases from Anki users' dataset HOT 9
- visualize metrics over time HOT 2
- [Feature Request] Train a gradient-boosted decision tree HOT 36
- Some weird first forgetting curves HOT 11
- [Feature request] Add confidence intervals for all metrics HOT 9
- accidental post
- Revlogs parsing HOT 12
- [Question] A βrawβ version of the tiny_dataset.zip HOT 3
- [Feature Request] Add a BiLSTM HOT 2
- [Feature request] Add the ACT-R model (see paper) HOT 21
- [TODO] Add DASH and its variants HOT 13
- [Feature request] A quantitative measure of cheating HOT 9
- Write an article about binned RMSE and cheating calibration metrics HOT 7
- Ebisu? HOT 7
- [Question] Some more details from a ML perspective HOT 8
- Cannot download dataset from huggingface HOT 4
- Neural network scheduler HOT 42
- Add MCC
Recommend Projects
-
React
A declarative, efficient, and flexible JavaScript library for building user interfaces.
-
Vue.js
π Vue.js is a progressive, incrementally-adoptable JavaScript framework for building UI on the web.
-
Typescript
TypeScript is a superset of JavaScript that compiles to clean JavaScript output.
-
TensorFlow
An Open Source Machine Learning Framework for Everyone
-
Django
The Web framework for perfectionists with deadlines.
-
Laravel
A PHP framework for web artisans
-
D3
Bring data to life with SVG, Canvas and HTML. πππ
-
Recommend Topics
-
javascript
JavaScript (JS) is a lightweight interpreted programming language with first-class functions.
-
web
Some thing interesting about web. New door for the world.
-
server
A server is a program made to process requests and deliver data to clients.
-
Machine learning
Machine learning is a way of modeling and interpreting data that allows a piece of software to respond intelligently.
-
Visualization
Some thing interesting about visualization, use data art
-
Game
Some thing interesting about game, make everyone happy.
Recommend Org
-
Facebook
We are working to build community through open source technology. NB: members must have two-factor auth.
-
Microsoft
Open source projects and samples from Microsoft.
-
Google
Google β€οΈ Open Source for everyone.
-
Alibaba
Alibaba Open Source for everyone
-
D3
Data-Driven Documents codes.
-
Tencent
China tencent open source team.
from srs-benchmark.