Comments (4)
Hi, the nn.L1Loss is the same as mean absolute error (MAE). For autoencoder, it is not clear in the paper but using MSE or MAE depends on the datasets. For this demo script, we use MSE for UCLA dataset. In general these two losses functions are pretty similar so I don't think there is huge difference between L1 and L2. it is also one line code that you could easily change to the other one.
from predict-cluster.
Thank you for the answer. And after training you demo script without changing any code, i found
the accuracy rate about 82% in the first few epochs, but in the subsequent training, the loss was reduced, and the accuracy of classification was in a downward trend.When the training is end in the 500th epoch, the accuracy of knn classification was only about 78%. so i want to know if have you meet the same problem, or if i should train more epochs to see the trend?
from predict-cluster.
oh you don't have to train 500 epochs. UCLA dataset is pretty small so it converges very fast. If you train too long, the model may be overfitting and the scores shown in the outputs are testing acc, so it is possible to decrease. And since the losses are designed to do the reconstruction task (instead of classification), decreasing loss does not necessary lead to better classification accuracy.
from predict-cluster.
oh OK thank you very much!!!
from predict-cluster.
Related Issues (20)
- Clarification re. Fixed Weight (FW) implementation HOT 3
- eval of each person for ntu dataset HOT 3
- About UWA3DII dataset HOT 2
- UWA dataset HOT 2
- pytorch implementation HOT 8
- Pretrained model on NTU-CS HOT 1
- Rotation Matrix R HOT 2
- error with get_feature() when run the train.py HOT 2
- some question on KNN HOT 2
- what's the shape of data at every HOT 2
- Train on 2D skeleton dataset HOT 4
- UWA3D handling HOT 1
- Action to predict HOT 5
- UCLA Data HOT 2
- About encoder states trajectories visualization HOT 2
- About problem in running ucla_demo HOT 1
- Your method cannot be called unsupervised!
- Dimension of the hidden layer HOT 4
- About NTU datasets HOT 34
Recommend Projects
-
React
A declarative, efficient, and flexible JavaScript library for building user interfaces.
-
Vue.js
🖖 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. 📊📈🎉
-
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 ❤️ Open Source for everyone.
-
Alibaba
Alibaba Open Source for everyone
-
D3
Data-Driven Documents codes.
-
Tencent
China tencent open source team.
from predict-cluster.