@@ -70,10 +70,10 @@ <h3>Large-scale ready</h3>
7070 < path d ="M5 15a7 7 0 0 0 12.2 4.8 "> </ path >
7171 </ svg >
7272 </ span >
73- < h3 > Inverse problem solvers</ h3 >
73+ < h3 > State-of-the-art solvers</ h3 >
7474 < p >
7575 Includes iterative methods for least-squares and
76- sparsity-promoting inversion workflows .
76+ proximal optimization .
7777 </ p >
7878 </ article >
7979 < article class ="feature-card ">
@@ -83,10 +83,11 @@ <h3>Inverse problem solvers</h3>
8383 < path d ="M12 4v2M12 18v2M4 12h2M18 12h2M6.4 6.4l1.4 1.4M16.2 16.2l1.4 1.4M17.6 6.4l-1.4 1.4M7.8 16.2l-1.4 1.4 "> </ path >
8484 </ svg >
8585 </ span >
86- < h3 > Stand-alone LinearOperator </ h3 >
86+ < h3 > Extensible base operators </ h3 >
8787 < p >
88- Provides a dedicated < code > LinearOperator</ code > interface
89- inspired by SciPy and tailored for PyLops usage.
88+ Provides < code > LinearOperator</ code > and
89+ < code > ProxOperator</ code > base classes inspired
90+ by SciPy and easy to extend.
9091 </ p >
9192 </ article >
9293 < article class ="feature-card ">
@@ -100,8 +101,8 @@ <h3>Stand-alone LinearOperator</h3>
100101 </ span >
101102 < h3 > Backend-agnostic design</ h3 >
102103 < p >
103- Uses an idiomatic approach that fits modern scientific Python
104- ecosystems and multiple computational backends.
104+ Uses an idiomatic approach to provide support
105+ for multiple computational backends (NumPy / CuPy / JAX) .
105106 </ p >
106107 </ article >
107108 < article class ="feature-card ">
@@ -111,7 +112,7 @@ <h3>Backend-agnostic design</h3>
111112 < path d ="M4 12h16M12 4a12 12 0 0 1 0 16M12 4a12 12 0 0 0 0 16 "> </ path >
112113 </ svg >
113114 </ span >
114- < h3 > Open-source project </ h3 >
115+ < h3 > Open-source driven </ h3 >
115116 < p >
116117 Community-driven and affiliated with NumFOCUS, with strong focus
117118 on transparent development.
0 commit comments