minor

README.md

@@ -54,6 +54,7 @@ Where each group supports:
   **`adjoint()`**, **`apply()`**, **`multiply()`**, **`inverse()`**,
   **`identity()`**, **`from_matrix()`**, and **`as_matrix()`** operations. (see
   [./examples/se3_example.py](./examples/se3_basics.py))
+- Taylor approximations near singularities.
 - Helpers for optimization on manifolds (see
   [./examples/se3_optimization.py](./examples/se3_optimization.py),
   <code>jaxlie.<strong>manifold.\*</strong></code>).
@@ -78,24 +79,10 @@ pip install jaxlie
 
 ---
 
-### In the wild
-
-- [jaxfg](https://github.com/brentyi/jaxfg) applies `jaxlie` to nonlinear least
-  squares problems with block-sparse structure. (for pose graph optimization,
-  bundle adjustment, etc)
-- [tensorf-jax](https://github.com/brentyi/tensorf-jax) is an unofficial
-  implementation of
-  [Tensorial Radiance Fields (Chen et al, ECCV 2022)](https://apchenstu.github.io/TensoRF/)
-  using `jaxlie`.
-  ![Render of a lego](https://github.com/brentyi/tensorf-jax/raw/main/lego_render.gif)
-
----
-
 ### Misc
 
-`jaxlie` was originally written for our IROS 2021 paper
-([link](https://github.com/brentyi/dfgo)). If it's useful for you, you're
-welcome to cite:
+`jaxlie` was originally written when I was learning about Lie groups for our IROS 2021 paper
+([link](https://github.com/brentyi/dfgo)):
 
 ```
 @inproceedings{yi2021iros,

jaxlie/_se2.py

@@ -138,7 +138,7 @@ def exp(cls, tangent: hints.Array) -> "SE2":
             jax.Array,
             jnp.where(
                 use_taylor,
-                jnp.ones_like(theta),  # Any non-zero value should do here.
+                1.0,  # Any non-zero value should do here.
                 theta,
             ),
         )
@@ -193,7 +193,7 @@ def log(self) -> jax.Array:
         # reverse-mode AD.
         safe_cos_minus_one = jnp.where(
             use_taylor,
-            jnp.ones_like(cos_minus_one),  # Any non-zero value should do here.
+            1.0,  # Any non-zero value should do here.
             cos_minus_one,
         )
 

jaxlie/_se3.py

@@ -131,7 +131,7 @@ def exp(cls, tangent: hints.Array) -> SE3:
             jax.Array,
             jnp.where(
                 use_taylor,
-                jnp.ones_like(theta_squared),  # Any non-zero value should do here.
+                1.0,  # Any non-zero value should do here.
                 theta_squared,
             ),
         )
@@ -173,7 +173,7 @@ def log(self) -> jax.Array:
         # reverse-mode AD.
         theta_squared_safe = jnp.where(
             use_taylor,
-            jnp.ones_like(theta_squared),  # Any non-zero value should do here.
+            1.0,  # Any non-zero value should do here.
             theta_squared,
         )
         del theta_squared

jaxlie/_so3.py

@@ -385,7 +385,7 @@ def exp(cls, tangent: hints.Array) -> SO3:
         safe_theta = jnp.sqrt(
             jnp.where(
                 use_taylor,
-                jnp.ones_like(theta_squared),  # Any constant value should do here.
+                1.0,  # Any constant value should do here.
                 theta_squared,
             )
         )