What is a Lie Group in layman's terms?

I'm having trouble getting my head arround the concept. Can someone explain it to me?

I think that understanding comes through examples. The most fundamental example I believe to be the rotation group. Consider the sphere $S^2\subset \mathbb{R}^3$. The sphere has rotational symmetries. If we rotate the sphere by any angle, the sphere doesn't change.

The collection of all rotations forms a Lie group. The group property basically means that if we rotate the sphere over any angle $\alpha$, after this over an angle $\beta$, it is the same if we would have rotated it in one go (over some different angle). Also any rotation has an inverse (rotating it over the opposite angle). This makes the rotations a group. The "Lie" in Lie group means that these rotations can be done arbitrary small. Many small rotations makes for a big rotation.

Lie groups capture the concept of "continuous symmetries".

Consider the set of $(n\times n)$ matrices that have non-zero determinant. Such a matrix corresponds to a system of linear equations ($n$ equations in $n$ unknowns) that has a unique solution. You can think of the solution as the unique point of intersection between the graph of a function and a horizontal hyperplane. Here it is helpful to think of $n=1$. In other words, the coefficients of the system correspond to a transformation of space: the variables $x_1, \ldots x_n$ are transformed to $\sum a_{ij} x_i$. The set of such transformations form a group: the matrices can be multiplied, each has an inverse, the multiplication is associative, and the identity transformation fixes each point of space.

Intuitively, it is easy to see which transformations are close to one another. They are close if they move points that are nearby to points that are nearby. Arithmetically, if the entries in the matrix are close, then the transformations are close: thus $0.14x + .33y$, is a reasonable approximation to $x/7+y/3$.

Thus the set of invertible $(n\times n)$ matrices is a space of invertible $(n\times n)$ matrices. What is not easy for a layman to see is that its spacial characteristics are defined via the determinant since as a set, the $(n\times n)$ matrices are a subset of $n^2$-space. The non-singular matrices are the pull-back of a regular value of the determinant function. [There is a small lie here: this is true for for matrices of determinant 1, but all non-zero determinant matrices deform onto that smaller space].

One important spacial characteristic is that these matrices form a smooth manifold. This is something that is analogous to the surface of a sphere (which is NOT a lie group), the surface of a torus (which is) or the $3$-dimensional sphere that consists of the set of $(x,y,z,w)$ such that $x^2+y^2+z^2+w^2=1$ (which also happens to be a Lie group).

From these examples, we abstract the idea of a Lie group which is a group (that can be thought of as a set of transformations or symmetries) that has the structure of a smooth manifold --- at small scales it is indistinguishable from ordinary Euclidean space. The multiplication and inversion maps are a differentiable functions. And these multiplications occur between pairs of symmetries --- they should not be confused with the action of the matrices on the vector space which is where I started the discussion.

Examples include the real line, the non-zero real numbers, the circle, the torus, the $3$-sphere, the set of rotations of 3-dimensional space, and the special unitary groups representations of which determine particles in physics.

There are some small problems with the definition that I gave. A smooth manifold is a topological space which is paracompact and Hausdorff (neither definition will play a role in the layman's understanding), and that is covered by coordinate charts with specific properties. I imagine that Wikipedia has the relevant definitions articulated carefully.