Prediction details
Each row in the following table shows the accuracy of predictions by an algorithm for individual cells in the data set. "Overall best CC" indicates the performance of the current best algorithm for that cell.
3. V1 Natural Image Sequences - List cells - All predictions - My predictions - Download data set - Submit prediction!| Algorithm: | TheMinimalist v1.0 |
| User: | pmineault |
| Notes: | I've grown frustrated trying to add fanciness to boosted models, because they're saturated with parameters and hence every time you add some useful nonlinearity you're eating up degrees of freedom that you would otherwise use for your spatial kernel. This is my attempt to create the lowest dimensional model that performs competitively with boosted models. Since the model is far from saturated in terms of parameters, hopefully with a couple of nonlinearities here and there it could beat r = 0.5. Involves 8 Gabors, a linear time filter, and some secret sauce for optimal deliciousness. Spatial filter has 15 free parameters, time filter 10, plus one gain. |
| Eval'd on: | 2011-03-15 05:13:37 (Prediction evaluated successfully) |
| Cell ID | My CC | Overall Best CC | |
| e0008 | 0.560 | 0.717 | |
| e0012 | 0.218 | 0.352 | |
| r0212b | 0.239 | 0.223 | |
| r0221a | 0.242 | 0.463 | |
| r0225c | 0.199 | 0.256 | |
| r0260 | 0.599 | 0.605 | |
| r0279 | 0.521 | 0.614 | |
| r0284 | 0.522 | 0.672 | |
| r0301 | 0.659 | 0.736 | |
| r0305 | 0.530 | 0.695 |
Not logged in. Log in - Create an account