-
Hello! I was wondering if it's possible for one to write their own `kernel_fn` in neural tangents to do regular kernel regression (for example, a Gaussian kernel). Naively just writing a `new_kernel_f…
-
I have built a custom layer (`KerasLayer`) using `class` in python (say `class NewLayer`). Can I use something like `stax.NewLayer` for manipulating `neural-tangents` on this custom layer?
-
I'm trying to differentiate through `predict_fn` provided by https://github.com/google/neural-tangents. This is doable with `jax.grad`, but not with `eagerpy.value_and_grad`.
ybj14 updated
4 years ago
-
Hi developers, I've met a problem when using neural-tangents as follows:
```
KERNEL_FN = nt.utils.batch.batch(KERNEL_FN, batch_size=kernel_batch_size)
AttributeError: module 'neural_tangents' has…
-
Hi, nice work, and thanks for sharing the code. When I was running the code, we encountered the following error.
```shell
jax._src.traceback_util.UnfilteredStackTrace: AttributeError: 'ShapedArra…
-
I am observing an error message when providing masked inputs with more than one feature dimensions to a kernel that involves `stax.GlobalAvgPool()`
Reproducer:
```
import jax
import jax.numpy as…
-
When I use neural-tangents 0.3.2 to run the 1d_regression.ipynb, I got ValuError:
Co…
-
Hey!
I was reading through the code and I noticed that you're using element-wise exponential matrix here:
https://github.com/google/neural-tangents/blob/5f286b7696364217aa4a2d92378aabd0203a791e/n…
-
I'm curious about the initialization for CNTK, so I replace the `kernel_fn` in `c_map(W_var, b_var)` function in [colab](https://colab.research.google.com/github/google/neural-tangents/blob/main/noteb…
-
## 一言でいうと
幅が無限とした場合のDNNを最急降下法で学習させる処理は、線形変換と同等であるとした論文(=学習結果は一次のテイラー展開で置き換えられる)。損失関数が2次の場合は出力がGaussianで維持されるため学習の過程はGaussian Processと見做すことができ、この点がBayesianNetと異なるとしている。
### 論文リンク
https://arxiv…