https://github.com/dragen1860/network2
https://github.com/dragen1860/network2
Last synced: 3 months ago
JSON representation
- Host: GitHub
- URL: https://github.com/dragen1860/network2
- Owner: dragen1860
- Created: 2018-01-21T06:01:19.000Z (over 7 years ago)
- Default Branch: master
- Last Pushed: 2018-04-14T07:39:58.000Z (over 7 years ago)
- Last Synced: 2025-04-02T20:14:57.912Z (7 months ago)
- Language: Python
- Homepage:
- Size: 74.3 MB
- Stars: 1
- Watchers: 3
- Forks: 0
- Open Issues: 0
-
Metadata Files:
- Readme: Readme.md
Awesome Lists containing this project
README
# Ominiglot
| Model | Fine Tune | 5-way Acc. | | 20-way Acc | |
|------------------------------------- |----------- |--------------- |--------------- |-------------- |--------------- |
| | | 1-shot | 5-shot | 1-shot | 5-shot |
| MANN | N | 82.8% | 94.9% | - | - |
| Matching Nets | N | 98.1% | 98.9% | 93.8% | 98.5% |
| Matching Nets | Y | 97.9% | 98.7% | 93.5% | 98.7% |
| MAML | Y | 98.7+-0.4% | 99.9+-0.1% | 95.8+-0.3% | 98.9+-0.2% |
| Meta-SGD | | 99.53+-0.26% | 99.93+-0.09% | 95.93+-0.38% | 98.97+-0.19% |
| TCML | | 98.96+-0.20% | 99.75+-0.11% | 97.64+-0.30% | 99.36+-0.18% |
| Learning to Compare | N | 99.6+-0.2% | 99.8+-0.1% | 97.6+-0.2% | 99.1+-0.1% |
| Ours(Res18, flatten features, 2 fc) | Y | 98.99% ,48120ep, 20b | 99.6%,48620ep, 5b | 96.99%,153920ep,20b | 97.2%, 63220ep,2b |
|naive, omni.py | N | 99.6% | 99.8% | 98.88% | |
|naive, omni.py, 2days | N | 99.80% | 100% | 98.88% | |
# mini-Imagenet
## Deeper version
| Model | Fine Tune | 5-way Acc. | | 20-way Acc | |
|-------------------------------------|-----------|------------|--------|------------|--------|
| | | 1-shot | 5-shot | 1-shot | 5-shot |
| Matching Nets | N | 43.56% | 55.31% | 17.31% | 22.69% |
| Meta-LSTM | | 43.44% | 60.60% | 16.70% | 26.06% |
| MAML | Y | 48.7% | 63.11% | 16.49% | 19.29% |
| Meta-SGD | | 50.49% | 64.03% | 17.56% | 28.92% |
| TCML | | 55.71% | 68.88% | - | - |
| Learning to Compare | N | 57.02% | 71.07% | - | - |
| ***reproduction* | N | 55.2% | 68.8% | | |
| ***Ours* | N | 53.0% | 64.6% | | |
## Naive version
| Model | Fine Tune | 5-way Acc. | | 20-way Acc | |
|-------------------------------------|-----------|------------|--------|------------|--------|
| | | 1-shot | 5-shot | 1-shot | 5-shot |
| Matching Nets | N | 43.56% | 55.31% | 17.31% | 22.69% |
| Meta-LSTM | | 43.44% | 60.60% | 16.70% | 26.06% |
| MAML | Y | 48.7% | 63.11% | 16.49% | 19.29% |
| Meta-SGD | | 50.49% | 64.03% | 17.56% | 28.92% |
| Learing to compare | N | 51.38% |67.07%| - | - |
| naivern.py (naive version) | N | 53.8% | 67.5% | - | - |
| naivern.py (naive version, avg pool, 9e-4) | N | 56.0%->60.8%, 2days | 68.1% | - | - |
| naive5.py (naive version, avg pool, sum over features) | N | | 72.7| - | - |
| naivesum.py (naive version, avg pool, sum over features, concat all setsz after f) | N | | 70.8| - | - |
## Simplified Deep version
| Model | Fine Tune | 5-way Acc. | | 20-way Acc | |
|-------------------------------------|-----------|------------|--------|------------|--------|
| | | 1-shot | 5-shot | 1-shot | 5-shot |
| Matching Nets | N | 43.56% | 55.31% | 17.31% | 22.69% |
| Meta-LSTM | | 43.44% | 60.60% | 16.70% | 26.06% |
| MAML | Y | 48.7% | 63.11% | 16.49% | 19.29% |
| Meta-SGD | | 50.49% | 64.03% | 17.56% | 28.92% |
| TCML | | 55.71% | 68.88% | - | - |
| Learning to Compare | N | 57.02% | 71.07% | - | - |
| rn.py, 463bottleneck, 6x6, conv->maxpool | N | 53.3% | | - | - |
| simrn.py, 111basicneck | N | 55.4% | 66.7% | - | - |
* 6x6 spational relation seems good, meaning larger spatial will lead to better
* naive version overcome all, meaning simplier nn
* the conv after repnet will occupy very huge gpu memory, but only reduce it will not lead to good, 51% on 5-way-1shot
* avg pooling
* sum on feature can not converge
* g network can enlarge since it have more batch
* spatial d is very critical, small d lead to faster converge, however, when it read 50%, it read the roof. Large `d` will converge slow but
have higher possibilityes.
* reduce Naive 4 conv to 3 conv will not improve performance(55.7%), and reduce d of spatial rn will not improve as well().
* add BatchNorm1d in self.g is extremely slow, will cost 3x time compared with none BN.
* add 4conv to 6conv will not improve: 68.7%
* smaller rn input, as 5x5, bad
* 12x12, no last layer avgpool, randomhorizontal flip, 65.9%
* deep version of relation, 64channel all, out 5x5 67.5%
>naivesum
0.712, 0.717, 0.736, 0.643, 0.672, 0.661, 0.653, 0.739, 0.677, 0.736, 0.741, 0.701, 0.603, 0.661, 0.693, 0.749, 0.709, 0.739, 0.659, 0.720, 0.688, 0.627, 0.717, 0.624, 0.715, 0.712, 0.741, 0.707, 0.704, 0.683, 0.683, 0.701, 0.728, 0.707, 0.715, 0.685, 0.720, 0.789, 0.680, 0.651, 0.664, 0.720, 0.784, 0.669, 0.744, 0.712, 0.600, 0.755, 0.717, 0.648, 0.595, 0.715, 0.731, 0.723, 0.680, 0.677, 0.699, 0.755, 0.688, 0.803, 0.707, 0.672, 0.651, 0.717, 0.747, 0.707, 0.744, 0.696, 0.712, 0.669, 0.717, 0.651, 0.704, 0.672, 0.696, 0.699, 0.645, 0.691, 0.768, 0.648, 0.685, 0.653, 0.616, 0.675, 0.629, 0.685, 0.736, 0.651, 0.691, 0.739, 0.635, 0.632, 0.707, 0.685, 0.693, 0.661, 0.643, 0.685, 0.699, 0.664, 0.773, 0.677, 0.709, 0.712, 0.659, 0.659, 0.733, 0.739, 0.787, 0.643, 0.747, 0.669, 0.643, 0.699, 0.733, 0.680, 0.680, 0.664, 0.704, 0.683,
accuracy: 0.694533333333 sem: 0.00744224200062
>slave2: original version, train for 24hours
0.707, 0.754, 0.689, 0.677, 0.677, 0.701, 0.704, 0.733, 0.738, 0.674, 0.677, 0.723, 0.717, 0.724, 0.708, 0.686, 0.702, 0.684, 0.674, 0.696, 0.744, 0.727, 0.793,
accuracy: 0.709146537842 sem: 0.0130249463636
>gpu: g+fc, f+fc
0.723, 0.750, 0.670, 0.693, 0.703, 0.707, 0.737, 0.727, 0.710, 0.673, 0.750, 0.673, 0.780, 0.650, 0.703, 0.623, 0.773, 0.823, 0.557, 0.693, 0.717, 0.720, 0.707, 0.667, 0.753,
accuracy: 0.707333333333 sem: 0.0222004602888
>p100:
0.627, 0.690, 0.637, 0.703, 0.700, 0.700, 0.667, 0.617, 0.690, 0.673, 0.723, 0.680, 0.720, 0.750, 0.697, 0.740, 0.713, 0.777, 0.693, 0.670, 0.743, 0.770, 0.730, 0.730, 0.713,
accuracy: 0.702133333333 sem: 0.0167383742684
>5way1shot fine tune from 60.8
0.553, 0.560, 0.687, 0.583, 0.560, 0.570, 0.610, 0.563, 0.523, 0.543, 0.560, 0.590, 0.603, 0.570, 0.620, 0.647, 0.610, 0.517, 0.493, 0.597, 0.507, 0.623, 0.640, 0.537, 0.617,
accuracy: 0.579333333333 sem: 0.01934070244
0.597, 0.720, 0.597, 0.557, 0.543, 0.473, 0.680, 0.640, 0.620, 0.640, 0.503, 0.533, 0.593, 0.507, 0.567, 0.583, 0.677, 0.637, 0.617, 0.507, 0.643, 0.647, 0.630, 0.497, 0.550,
accuracy: 0.590266666667 sem: 0.0268392648054
>metric
0.69067, 0.72000, 0.68533, 0.68800, 0.70133, 0.70933, 0.70933, 0.79200, 0.72267, 0.74400, 0.68533, 0.72533, 0.58667, 0.73867, 0.73067, 0.71200, 0.70400, 0.71200, 0.68800, 0.70667, 0.70933, 0.76533, 0.74400, 0.72267, 0.68267, 0.69333, 0.73067, 0.73067, 0.66133, 0.68800, 0.69600, 0.74133, 0.69600, 0.73333, 0.70667, 0.72000, 0.74400, 0.67200, 0.76533, 0.65333, 0.66133, 0.76533, 0.69067, 0.72533, 0.62933, 0.66133, 0.66667, 0.72267, 0.70933, 0.68800, 0.71467, 0.68533, 0.72000, 0.72000, 0.71733, 0.71200, 0.65600, 0.74400, 0.74933, 0.67200, 0.71733, 0.69067, 0.68800, 0.70667, 0.76800, 0.67733, 0.68267, 0.75733, 0.69867, 0.68800, 0.74667, 0.67733, 0.70933, 0.67200, 0.67200, 0.68800, 0.66933, 0.64267, 0.67200, 0.74667, 0.77333, 0.72533, 0.77067, 0.72000, 0.74133, 0.71200, 0.78400, 0.68533, 0.67200, 0.67467, 0.68533, 0.75467, 0.69867, 0.64533, 0.66400, 0.73067, 0.77067, 0.69333, 0.62667, 0.68000, 0.71467, 0.63733, 0.63733, 0.70133, 0.67200, 0.64267, 0.66933, 0.73600, 0.62933, 0.69600, 0.67200, 0.71200, 0.68000, 0.62933, 0.70933, 0.70933, 0.73867, 0.73067, 0.67733, 0.73333,
accuracy: 0.702711111111 sem: 0.00676532628843
<<<<<<<<< accuracy: 0.702711111111 best accuracy: 0.699428571429 >>>>>>>>
>learnign2compare: 5way5shot:
0.63111, 0.66222, 0.73333, 0.72444, 0.68889, 0.70222, 0.71111, 0.70222, 0.71556, 0.66222, 0.76889, 0.64000, 0.71111, 0.67111, 0.60889, 0.68889, 0.68444,
accuracy: 0.6886274509803921 sem: 0.02055769936837831
>slave2 5way 1shot:0.54133, 0.64533, 0.52533, 0.55733, 0.58400, 0.58133, 0.61067, 0.55467, 0.62400, 0.61867, 0.55733, 0.59200, 0.65333, 0.62667, 0.50667, 0.59200, 0.57067, 0.62133, 0.53867, 0.64533, 0.56800, 0.54400, 0.55733,
accuracy: 0.583304347826 sem: 0.0178551543546
>20way 1shot: 0.98333, 0.99667, 0.99333, 0.96667, 0.99667, 0.97667, 0.98000, 1.00000, 0.98333, 0.98667, 0.99667, 0.96667, 0.98000, 0.99667, 0.96333, 0.98000, 0.99667, 0.99667, 0.98667, 0.99333, 0.99333, 1.00000, 0.99000, 0.99000, 0.99000, 1.00000, 0.98667, 0.99667, 0.99333, 0.99333, 0.99667, 0.96000, 0.96667, 0.98667, 0.99333, 0.97333, 0.99667, 0.97333, 0.99333, 0.98667, 0.97667, 0.99667, 0.99667, 0.98333, 0.97667, 0.99333, 0.98333, 0.97000, 0.98667, 0.95667, 0.98000, 0.97333, 0.99667, 0.97333, 0.98667, 0.98000, 0.99667, 0.96000, 0.97000, 0.96667, 0.99333, 0.96667, 0.99000, 0.98667, 0.99000, 0.99000, 0.99333, 1.00000, 0.99000, 0.98333, 0.98000, 0.96333, 0.99000, 0.99333, 0.98667, 0.97000, 0.97333, 0.98333, 0.99000, 0.97000, 0.99667, 0.99667, 0.99333, 0.97667, 0.98000, 0.99667, 0.96667, 0.97667, 0.98000, 0.98000, 0.99000, 0.99667, 0.97000, 0.99000, 0.99333, 0.99333, 0.98667, 0.99333, 0.99333, 0.99333, 0.98667, 0.97333, 0.97667, 0.99333, 0.98667, 0.97667, 0.99000, 0.91000, 0.98667, 0.98000, 0.93333, 0.97667, 0.98333, 0.99333, 1.00000, 0.98333, 0.99667, 0.98000, 1.00000, 0.99333, 0.97667, 0.98000, 0.99333, 0.98667, 0.99000, 0.99333, 1.00000, 0.99667, 0.99000, 0.98333, 1.00000, 0.98000, 0.98667, 0.99333, 0.93667, 0.98333, 0.98000, 0.99000, 0.98667, 0.97667, 0.95000, 0.98667, 0.98667, 0.98000, 1.00000, 0.98000, 0.94667, 0.98667, 0.97333, 0.99333, 0.99667, 0.99000, 0.98667, 0.98000, 0.99667, 0.99000, 0.97667, 0.99333, 0.97667, 1.00000, 0.97000, 0.98333, 0.97333, 0.99000, 0.95333, 0.97667, 0.94333, 0.99000, 0.99000, 0.99333, 0.98000, 0.99667, 0.95000, 0.99000, 0.98333, 0.96667, 0.99333, 0.96333, 0.99000, 0.98333, 0.98333, 0.99000, 1.00000, 0.99000, 0.99667, 0.99000, 0.98667, 0.99333, 0.97667, 0.96000, 0.97667, 1.00000, 0.99333, 0.99000, 0.98667, 0.98667, 0.97333, 0.98333, 0.99333, 0.99333, 0.99000, 0.98333, 0.93667, 0.99333, 0.99000, 0.98667, 0.97000, 0.98667, 0.97667, 0.99333, 0.99667, 0.99667, 0.99000, 0.99000, 0.97667, 0.99333, 0.95333, 0.98333, 0.99000, 0.99333, 0.94000, 0.99000, 0.99333, 0.98667, 0.98333, 1.00000, 0.99667, 0.99000, 0.98667, 0.97333, 0.97333, 0.96000, 0.99667, 0.99333, 0.97333, 0.99333, 1.00000, 0.99000, 0.97000, 0.98667, 0.99667, 0.93333, 0.99667, 0.98333, 0.98000, 0.96333, 0.95333, 0.99333, 0.96000, 0.97000, 0.99333, 0.95667, 1.00000, 0.98333, 0.99667, 0.99333, 0.97667, 0.98667, 0.99333, 0.96000, 0.98000, 1.00000, 0.97000, 0.99000, 0.94667, 0.92333, 0.98333, 0.99667, 0.97000, 0.99667, 1.00000, 0.99333, 0.97667, 0.98000, 0.99000, 0.98333, 0.99333, 0.99667, 0.99667, 0.99667, 0.96333, 0.99667, c0.97333, 0.96333, 0.93667, 0.98000, 0.98000, 0.99000, 0.98333, 0.99667, 0.92000, 0.98667, 0.94667, 0.98667, 0.97000, 0.97333, 0.98667, 0.97000, 0.99667, 0.99333, 0.99000, 0.99667, 1.00000, 0.98000, 0.99000, 0.99667, 0.99667, 0.96333, 0.89667, 0.93333, 0.99667, 1.00000, 0.98667, 0.99333, 0.98000, 0.97000, 0.99667, 0.99333, 0.99000, 0.98667, 0.99333, 0.99667, 0.99667, 0.98000, 0.99667, 0.99000, 0.97000, 0.99000, 0.98667, 0.99667, 0.99333, 0.99000, 0.99667, 0.98333, 0.98000, 0.97000, 0.96333, 0.97000, 0.98333, 0.99000, 0.99333, 0.99667, 0.98667, 1.00000, 0.97333, 0.98667, 0.98333, 0.99000, 0.99333, 0.98000, 0.98333, 0.97667, 0.99667, 0.93333, 0.99333, 0.95667, 0.99333, 0.97667, 0.99333, 0.99333, 0.93333, 0.98667, 0.98667, 0.96000, 0.99667, 0.97333, 0.99000, 0.97333, 0.91667, 0.99000, 0.98667, 0.96333, 0.95333, 1.00000, 0.93667, 0.99000, 1.00000, 0.95000, 0.99333, 0.98000, 0.99333, 0.94333, 0.98333, 0.99000, 0.97667, 0.98667, 0.97667, 0.98667, 0.99333, 0.98000, 0.99333, 0.96667, 0.99333, 0.96000, 1.00000, 0.97667, 0.96000, 0.96333, 0.99333, 1.00000, 0.97000, 0.99000, 0.98333, 0.99333, 0.98333, 0.97000, 0.97333, 0.99000, 0.96667, 0.99333, 0.98667, 0.97667, 0.99333, 0.98000, 0.99000, 0.99333, 0.99333, 0.97667, 0.93000, 0.95000, 0.97333, 1.00000, 1.00000, 0.99667, 1.00000, 0.97667, 0.99333, 0.97333, 0.99000, 1.00000, 0.98333, 1.00000, 0.99333, 0.99333, 0.96667, 1.00000, 0.97667, 0.94667, 0.98333, 0.98333, 0.98000, 0.97333, 0.99000, 0.99667, 0.99000, 0.99667, 0.97333, 0.95667, 0.98333, 0.98333, 0.98333, 0.99000, 0.99000, 0.97333, 0.99000, 0.98667, 0.93333, 0.99667, 0.97000, 0.99333, 0.98333, 0.98667, 0.99333, 0.98333, 0.99333, 0.98000, 0.93667, 0.98333, 0.98333, 0.99333, 0.96333, 0.98000, 0.99000, 0.94333, 0.98333, 0.99000, 0.98667, 0.99667, 0.97000, 0.97000, 0.98667, 0.97667, 0.99000, 0.94667, 0.98000, 0.98000, 0.97000, 0.99000, 0.99667, 0.97667, 0.98333, 0.99333, 0.97333, 0.99333, 0.99000, 0.95667, 0.99667, 0.93333, 0.98667, 0.97667, 0.94333, 0.99333, 0.99667, 0.99667, 0.95000, 0.98000, 0.99000, 0.96333, 0.94333, 0.98667, 0.97333, 0.99000, 0.97667, 1.00000, 0.99000, 0.97667, 1.00000, 0.99667, 0.97333, 0.95667, 0.98667, 0.98333, 0.98333, 0.99333, 0.97667, 0.99667, 0.97667, 0.96667, 0.99000, 0.99333, 0.99000, 0.99000, 0.99667, 0.97667, 0.98000, 0.98667, 0.98667, 0.95000, 0.98333, 0.95667, 0.97333, 0.98667, 1.00000, 0.99000, 0.96333, 0.99667, 0.92000, 0.99667, 0.99000, 0.99333, 0.94000, 0.99000, 1.00000, 0.97333, 0.98667, 0.99000, 0.99333, 0.97000, 1.00000, 0.99333, 0.99333, 0.98333, 0.98333, 0.99000, 0.96667, 0.99667, 0.99333, 0.99667, 0.95333, 0.99333, 0.94333, 0.99333, 0.99000, 0.99000, 1.00000, 0.96000, 0.96000, 0.99667, 0.99667, 0.99333, 0.96333, 0.98000, 0.96000, 0.96333, 0.98667, 0.98667, 0.99333, 0.98667, 0.99333, 0.99000, 0.94333, 0.99667, 0.99333, 0.99333, 0.94000, 0.94667, 0.99667, 0.96333, 0.97000, 0.99667,
accuracy: 0.981938888889 sem: 0.00131637273997
>20way5shot:
0.99667, 0.99667, 0.99889, 0.99889, 0.99667, 0.99111, 0.99778, 0.99556, 0.99778, 0.99778, 0.99667, 0.99778, 0.99444, 0.99222, 0.98556, 0.99000, 1.00000, 0.99889, 0.99444, 0.99556, 0.98667, 0.98556, 0.98222, 0.99556, 0.99778, 0.99778, 0.99444, 0.99889, 0.99111, 0.99111, 1.00000, 1.00000, 0.99778, 0.99667, 0.99667, 0.99222, 0.99556, 0.99444, 0.99333, 0.99444, 0.99778, 0.99778, 1.00000, 1.00000, 0.99444, 0.99222, 1.00000, 0.99778, 0.99556, 0.98556, 1.00000, 0.99556, 0.99778, 0.98444, 0.99222, 0.99333, 0.99444, 0.99667, 0.98889, 0.98333, 0.99000, 0.99889, 0.99778, 0.99778, 0.98667, 0.99444, 0.99333, 0.99444, 0.99778, 0.99444, 0.99111, 0.99556, 0.99889, 0.99222, 0.99667, 0.99444, 0.99667, 0.99333, 0.99667, 0.99778, 0.99667, 0.98889, 0.99556, 0.99222, 0.99111, 0.98111, 0.99222, 0.99556, 0.98111, 0.98111, 0.99444, 0.99667, 0.99111, 0.99667, 0.99667, 0.99778, 0.99556, 0.99667, 0.99444, 0.99111, 0.99667, 0.98667, 0.99444, 0.98222, 0.99667, 0.99889, 0.99444, 0.99667, 0.99889, 0.99333, 0.99778, 0.98778, 0.99556, 0.99444, 0.99667, 0.99667, 0.99444, 0.99222, 0.97556, 0.98667, 0.99667, 0.99556, 0.97667, 0.99556, 0.99556, 0.99333, 0.99778, 0.98889, 0.99333, 0.99556, 0.98667, 0.99667, 0.99667, 0.99667, 0.99333, 0.99889, 0.99889, 0.99667, 0.99333, 0.99778, 0.99333, 1.00000, 0.99444, 0.99889, 0.99778, 0.99444, 0.99222, 0.99556, 0.99667, 0.99667, 0.99889, 0.99778, 0.99333, 0.99556, 0.99778, 0.98111, 0.99667, 0.99444, 0.99667, 0.99667, 0.97556, 0.99556, 0.99778, 0.99222, 0.99222, 0.99889, 0.99778, 0.99444, 0.99667, 0.99778, 0.99556, 0.99444, 0.99444, 0.99889, 0.99222, 0.99667, 0.99556, 0.99444, 0.99556, 0.98889, 0.99667, 0.99222, 0.99556, 0.98111, 0.99556, 0.99889, 0.99667, 0.99333, 0.99667, 0.99444, 0.98667, 0.99222, 0.98778, 0.98778, 0.99444, 0.98889, 0.99889, 0.99778, 0.98778, 0.99444,
accuracy: 0.994111111111 sem: 0.000670760856559
0.99400, 0.98667, 0.99867, 0.99333, 0.99600, 0.99333, 0.99733, 0.99933, 0.99400, 0.99533, 0.99600, 0.99800, 0.99333, 0.99733, 0.99400, 0.99533, 0.99533, 0.99533, 0.99533, 0.99800, 0.99733, 0.99600, 0.99733, 0.99333, 0.99667, 0.99533, 0.98867, 0.98933, 0.99600, 0.99000, 0.99400, 0.99467, 0.99067, 0.99533, 0.99400, 0.99600, 0.99200, 0.99533, 0.99733, 0.99333, 0.99800, 0.99667, 0.99400, 0.99733, 0.99267, 0.99533, 0.99667, 0.99133, 0.98400, 0.99200, 0.99333, 0.99533, 0.99467, 0.99800, 0.99200, 0.99533, 0.99867, 0.99467, 0.99733, 0.99600, 0.99933, 0.99267, 0.99600, 0.99667, 0.99733, 0.99600, 0.99467, 0.99533, 0.99800, 0.99667, 0.99667, 0.99733, 0.99533, 0.99867, 0.99733, 0.99867, 0.99267, 0.99067, 0.98933, 0.99667, 0.99533, 0.99000, 0.99133, 0.99133, 0.99067, 0.99467, 0.99800, 0.99267, 0.99800, 0.99867, 0.99267, 0.99533, 0.99533, 0.99600, 0.98600, 0.99200, 0.98733, 0.99467, 0.99333, 0.99067, 0.99800, 0.99800, 0.98400, 0.99400, 0.99333, 0.99800, 0.99533, 0.99733, 0.99533, 0.99733, 0.99867, 0.99600, 0.99800, 0.99667, 0.99333, 0.99200, 0.99667, 0.99733, 0.99467, 0.99600,
accuracy: 0.994766666667 sem: 0.000560808026472
>5way5shot omniglot:
1.00000, 1.00000, 1.00000, 1.00000, 1.00000, 0.99667, 1.00000, 1.00000, 1.00000, 1.00000, 1.00000, 1.00000, 1.00000, 1.00000, 1.00000, 1.00000, 1.00000, 1.00000, 0.99667, 1.00000, 1.00000, 1.00000, 0.99333, 1.00000, 0.99667, 1.00000, 1.00000, 0.99667, 1.00000, 1.00000, 1.00000, 1.00000, 1.00000, 1.00000, 1.00000, 0.99667, 1.00000, 0.99000, 1.00000, 1.00000, 1.00000, 1.00000, 1.00000, 1.00000, 0.99333, 1.00000, 1.00000, 1.00000, 1.00000, 1.00000, 1.00000, 1.00000, 1.00000, 0.99667, 1.00000, 0.99667, 0.99667, 1.00000, 1.00000, 1.00000, 1.00000, 1.00000, 1.00000, 1.00000, 0.99667, 1.00000, 1.00000, 1.00000, 1.00000, 0.99667, 1.00000, 1.00000, 1.00000, 0.99333, 1.00000, 1.00000, 1.00000, 1.00000, 1.00000, 0.99667, 0.99667, 1.00000, 0.99667, 1.00000, 1.00000, 0.99667, 1.00000, 0.99000, 1.00000, 1.00000, 0.99000, 1.00000, 1.00000, 1.00000, 1.00000, 1.00000, 1.00000, 1.00000, 1.00000, 1.00000, 1.00000, 0.99667, 0.96667, 1.00000, 1.00000, 1.00000, 1.00000, 1.00000, 1.00000, 1.00000, 0.99667, 1.00000, 0.99667, 1.00000, 1.00000, 1.00000, 1.00000, 0.99333, 1.00000, 0.99667, 1.00000, 1.00000, 0.99333, 1.00000, 1.00000, 1.00000, 1.00000, 1.00000, 0.99000, 1.00000, 1.00000, 0.99667, 1.00000, 1.00000, 1.00000, 1.00000, 1.00000, 1.00000, 0.99667, 0.99667, 0.99333, 1.00000, 1.00000, 0.99333, 1.00000, 1.00000, 1.00000, 1.00000, 0.99667, 1.00000,
accuracy: 0.998711111111 sem: 0.000563489814168
>5way1shot omniglot:
>5way1shot omniglot
0.99000, 0.99833, 0.99667, 1.00000, 1.00000, 0.99500, 0.99833, 0.99500, 0.99833, 0.99833, 1.00000, 0.99667, 0.99667, 0.99833, 1.00000, 1.00000, 0.99833, 0.99833, 1.00000, 1.00000, 0.99667, 0.99167, 0.99667, 0.99833, 0.99333, 1.00000, 0.99833, 1.00000, 0.99667, 0.99667, 0.99667, 0.99833, 0.99833, 0.99500, 0.99833, 0.99500, 0.99667, 0.99833, 0.99833, 1.00000, 0.99167, 1.00000, 1.00000, 0.99667, 1.00000, 0.99667, 0.99833, 1.00000, 0.99833, 1.00000, 1.00000, 0.99833, 1.00000, 0.99833, 1.00000, 0.99667, 1.00000, 1.00000, 1.00000, 0.99833, 0.99667, 0.99333, 1.00000, 0.99833, 0.99833, 1.00000, 0.99833, 1.00000, 0.99667, 1.00000, 1.00000, 1.00000, 1.00000, 1.00000, 0.99833,
accuracy: 0.998066666667 sem: 0.000512305004666
>5wah1shot: imagenet:
0.52000, 0.66133, 0.57333, 0.59733, 0.55467, 0.58667, 0.47733, 0.56267, 0.56000, 0.52533, 0.56800, 0.53333, 0.61867, 0.62133, 0.57867, 0.54400, 0.57067, 0.55200, 0.62133, 0.64267, 0.50933, 0.58133, 0.60533, 0.55200, 0.58400, 0.59733, 0.48533, 0.54933, 0.57067, 0.57333, 0.61333, 0.52533, 0.64267, 0.59733, 0.56800, 0.54667, 0.49067, 0.55200, 0.59467, 0.57867, 0.55733, 0.58400, 0.58400, 0.48267, 0.62400, 0.57067, 0.62133, 0.56533, 0.56800, 0.55467, 0.57600, 0.54667, 0.62133, 0.56800, 0.51733, 0.57067, 0.61600, 0.57333, 0.66133, 0.50933, 0.58133, 0.60267, 0.52533, 0.60000, 0.63467, 0.52800, 0.55733, 0.58667, 0.58667, 0.61867, 0.56800, 0.54400, 0.59467, 0.57067, 0.56533, 0.60533, 0.55467, 0.63733, 0.58933, 0.59733, 0.59733, 0.59200, 0.56000, 0.60800, 0.57867, 0.59467, 0.58400, 0.53867, 0.50933, 0.59200, 0.57600, 0.57333, 0.63733, 0.52800, 0.56800, 0.56267, 0.65333, 0.62667, 0.51733, 0.54400, 0.51467, 0.56533, 0.53867, 0.50933, 0.53333, 0.54933, 0.58400, 0.60000, 0.56533, 0.54400, 0.52267, 0.59467, 0.58133, 0.56800, 0.53333, 0.60267, 0.54400, 0.56533, 0.59467, 0.57333,
accuracy: 0.571933333333 sem: 0.00687127231147
>20way1shot miniimage:
0.21667, 0.22333, 0.22667, 0.18333, 0.25000, 0.20000, 0.19333, 0.27333, 0.20000, 0.23000, 0.22333, 0.22000, 0.15000, 0.21667, 0.24667, 0.19000, 0.20333, 0.29667, 0.25667, 0.23333, 0.16667, 0.19667, 0.24000, 0.24667, 0.19000, 0.26333, 0.24667, 0.24667, 0.28333, 0.20000, 0.21667, 0.18667, 0.20333, 0.24333, 0.24667, 0.23667, 0.20667, 0.20667, 0.20000, 0.24667, 0.24333, 0.21667, 0.18333, 0.21667, 0.22000, 0.21667, 0.25667, 0.21000, 0.23333, 0.21000, 0.23667, 0.22667, 0.19667, 0.22000, 0.18667, 0.21333, 0.25667, 0.19333, 0.23667, 0.19333, 0.20667, 0.21000, 0.24333, 0.31000, 0.19667, 0.22000, 0.20667, 0.22667, 0.27000, 0.24000, 0.22333, 0.20667, 0.25667, 0.22333, 0.15667, 0.27333, 0.23333, 0.23000, 0.29000, 0.17333, 0.19333, 0.16333, 0.19000, 0.24667, 0.18000, 0.18333, 0.23667, 0.23667, 0.21333, 0.20333, 0.28000, 0.16333, 0.27333, 0.22333, 0.19333, 0.18333, 0.21333, 0.24000, 0.16000, 0.27667, 0.23000, 0.22333, 0.20667, 0.19667, 0.21667, 0.23333, 0.25667, 0.20333, 0.24667, 0.24333, 0.21667, 0.25667, 0.21333, 0.24333, 0.21333, 0.25667, 0.19667, 0.21000, 0.26000, 0.20000, 0.22000, 0.23333, 0.24000, 0.19667, 0.21667, 0.27667, 0.23000, 0.24667, 0.19667, 0.25667, 0.31333, 0.22000, 0.18667, 0.17667, 0.23333, 0.23000, 0.19667, 0.24333, 0.21667, 0.22000, 0.24000, 0.25667, 0.24667, 0.20667, 0.20333, 0.21000, 0.26333, 0.14000, 0.29000, 0.19000, 0.24667, 0.28333, 0.20000, 0.25667, 0.18667, 0.19667, 0.20333, 0.24000, 0.22333, 0.22333, 0.20333, 0.19667, 0.22667, 0.21333, 0.23667, 0.23000, 0.21333, 0.23333, 0.22333, 0.17333, 0.19000, 0.27000, 0.30333, 0.21000, 0.21667, 0.21333, 0.23667, 0.18667, 0.24333, 0.17667, 0.26667, 0.24000, 0.25000, 0.15667, 0.21000, 0.23000, 0.22000, 0.30333, 0.15667, 0.23333, 0.21667, 0.23333, 0.20333, 0.20333, 0.25667, 0.28667, 0.21000, 0.19667, 0.24000, 0.23333, 0.27000, 0.24667, 0.28667, 0.23667, 0.18333, 0.20667, 0.16000, 0.22000, 0.23333, 0.24667, 0.21667, 0.20667, 0.22667, 0.22667, 0.18333, 0.19000, 0.17667, 0.19667, 0.21667, 0.20333, 0.17333, 0.19000, 0.25000, 0.17333, 0.23333, 0.23333, 0.26333, 0.21333, 0.26000, 0.20000, 0.24333, 0.20667, 0.19667, 0.21333, 0.17667, 0.18000, 0.20333, 0.23667, 0.22333, 0.21667, 0.25333, 0.23333, 0.23000, 0.17667, 0.24333, 0.19667, 0.22667, 0.18333, 0.25333, 0.20667, 0.16333, 0.29333, 0.22667, 0.19333, 0.16667, 0.20000, 0.16667, 0.17000, 0.22667, 0.20000, 0.21000, 0.22333, 0.24000, 0.25333, 0.19333, 0.22667, 0.27333, 0.23333, 0.19333, 0.25000, 0.21000, 0.25333, 0.25000, 0.21333, 0.24333, 0.25333, 0.19333, 0.23667, 0.25667, 0.24667, 0.24333, 0.17667, 0.25000, 0.23667, 0.19667, 0.23000, 0.20667, 0.25000, 0.22667, 0.18000, 0.25667, 0.22333, 0.23000, 0.20333, 0.26667, 0.23333, 0.24333, 0.26000, 0.23667, 0.23000, 0.19000, 0.24000, 0.22333, 0.20000, 0.25000, 0.18333, 0.23000, 0.21000, 0.21667, 0.19333, 0.28667, 0.23000, 0.20000, 0.19333, 0.22333, 0.25667, 0.19667, 0.20667, 0.19333, 0.18667, 0.20000, 0.24667, 0.23667, 0.22667, 0.24667, 0.23000, 0.22333, 0.20000, 0.20667, 0.17000, 0.18667, 0.22667, 0.21333, 0.20333, 0.19000, 0.17333, 0.19000, 0.21667, 0.26667, 0.21333, 0.18667, 0.22667, 0.26000, 0.19667, 0.28000, 0.27333, 0.22000, 0.13667, 0.21333, 0.23333, 0.23333, 0.19000, 0.18333, 0.18000, 0.25667, 0.24667, 0.18000, 0.24000, 0.20667, 0.23000, 0.26000, 0.26333, 0.19000, 0.23667, 0.28667, 0.20333, 0.13333, 0.20333, 0.23333, 0.14000, 0.14000, 0.18667, 0.20000, 0.22333, 0.27667, 0.17333, 0.19333, 0.23333, 0.24667, 0.25667, 0.26000, 0.21000, 0.24667, 0.23333, 0.20667, 0.23333, 0.22667, 0.21000, 0.21000, 0.15667, 0.24667, 0.22000, 0.26333, 0.20667, 0.18333, 0.23000, 0.26000, 0.22333, 0.19333, 0.19333, 0.18333, 0.21333, 0.26333, 0.26000, 0.15000, 0.24667, 0.24667, 0.26000, 0.19000, 0.20000, 0.23000, 0.19667, 0.17667, 0.20000, 0.24667, 0.21333, 0.24667, 0.17000, 0.19667, 0.23333, 0.24333, 0.28000, 0.24667, 0.20667, 0.20000, 0.19667, 0.22333, 0.18667, 0.23333, 0.23333, 0.20667, 0.23667, 0.20667, 0.24333, 0.21000, 0.28000, 0.26000, 0.27667, 0.25000, 0.22667, 0.19000, 0.21000, 0.17667, 0.20333, 0.24000, 0.23333, 0.20000, 0.23333, 0.27333, 0.23333, 0.20333, 0.29333, 0.24000, 0.25667, 0.29667, 0.25333, 0.17667, 0.16667, 0.21000, 0.21667, 0.24333, 0.27667, 0.21000, 0.26000, 0.20000, 0.17667, 0.27667, 0.21000, 0.19667, 0.24000, 0.22667, 0.25333, 0.27667, 0.20333, 0.24333, 0.21333, 0.17000, 0.19667, 0.29000, 0.19667, 0.21667, 0.17000, 0.26667, 0.17667, 0.23667, 0.19333, 0.22667, 0.23667, 0.22333, 0.23333, 0.26333, 0.26333, 0.19000, 0.25333, 0.16667, 0.27667, 0.29000, 0.22667, 0.21667, 0.28333, 0.24000, 0.22000, 0.18333, 0.20333, 0.24667, 0.29333, 0.23333, 0.25333, 0.23000, 0.24667, 0.27667, 0.27000, 0.22333, 0.24333, 0.31333, 0.21000, 0.23333, 0.27333, 0.24667, 0.23667, 0.18000, 0.26667, 0.26667, 0.24000, 0.25333, 0.26333, 0.16333, 0.22667, 0.22667, 0.28333, 0.19000, 0.18000, 0.20333, 0.21333, 0.24000, 0.19333, 0.21333, 0.21667, 0.21333, 0.20000, 0.21667, 0.24333, 0.22000, 0.17667, 0.22667, 0.21667, 0.16667, 0.18000, 0.18667, 0.27333, 0.20667, 0.26333, 0.20000, 0.20333, 0.18667, 0.22667, 0.23000, 0.24333, 0.24333, 0.17000, 0.21000, 0.25000, 0.29333, 0.20000, 0.19333, 0.20667, 0.20000, 0.26667, 0.18333, 0.19333, 0.20000, 0.21333, 0.24667, 0.16333, 0.20333, 0.13333, 0.24333, 0.16000, 0.17333, 0.25000, 0.21000, 0.24667, 0.19667, 0.20000, 0.20000, 0.18333, 0.28000, 0.19000, 0.15333, 0.22333, 0.21333, 0.21667, 0.26000, 0.29667, 0.20667, 0.21000, 0.21333, 0.25667, 0.20000, 0.27667,
accuracy: 0.222011111111 sem: 0.002628248652