In [1]:

```
%pylab
%matplotlib inline
```

Check out the formula for a companion matrix on wikipedia.
Define a function `companion`

which accepts a vector in argument, and returns the corresponding companion matrix.
You can use the command `diag`

for that.
The resulting matrix should be of complex type.

In [60]:

```
diag?
```

In [ ]:

```
ones?
```

In [32]:

```
def companion(coefficients):
pass
```

In [62]:

```
C = companion(ones(3))
assert(C.dtype == complex)
assert(len(C) == 3)
```

Fix a given size, say `size = 20`

, and create a vector of length 20 with random, normally distributed, complex numbers. Use the `randn`

function for that, and combine two random real vectors to get a random complex vector.

In [ ]:

```
randn?
```

Now, fix a standard deviation, say `sigma = 1./10`

, and use the random complex coefficients obtained above, multiplied by sigma, in the `companion`

function. Use `eigvals`

to compute the eigenvalues, and plot them on the complex plane. You can use the command `axis('equal')`

to make sure the plot has the same dimensions in x and y.

Finally, repeat that, say 200 times. Plot all the eigenvalues on the same figure. What do you observe? What happens when you change the standard deviation `sigma`

?

In [1]:

```
size = 20
for i in range(200):
pass
```

In [ ]:

```
```