Numpy filter using condition on each element

Question

I have a filter expression as follows:

feasible_agents = filter(lambda agent: agent >= cost[task, agent], agents)

where agents is a python list.

Now, to get speedup, I am trying to implement this using numpy.

What would be the equivalent using numpy?

I know that this works:

threshold = 5.0
feasible_agents = np_agents[np_agents > threshold]

where np_agents is the numpy equivalent of agents.

However, I want threshold to be a function of each element in the numpy array.

"I want threshold to be a function of each element in the numpy array." Please explain. — cs95
– cs95, Commented Sep 18, 2018 at 7:22
As you can see from the first expression using lambda, the threshold is the cost value that is obtained by lookup in the cost table using each element of the agents list as key. task is a constant here. i.e. the threshold is not fixed, but will vary for every element agent in the numpy array np_agents. — Niloy Saha
– Niloy Saha, Commented Sep 18, 2018 at 7:26
Difficult to vectorize a solution involving a dictionary. Maybe provide a minimal reproducible example. If the keys to your dict are integers, something may be possible here. — cs95
– cs95, Commented Sep 18, 2018 at 7:36

unlut · Accepted Answer · 2018-09-18 07:28:08Z

15

You can use numpy.extract:

>>> nparr = np.array([1, 2, 3, 4, 5, 6, 7, 8, 9, 10])
>>> nparreven = np.extract(nparr % 2 == 0, nparr)

or numpy.where:

>>> nparr = np.array([1, 2, 3, 4, 5, 6, 7, 8, 9, 10])
>>> nparreven = nparr[np.where(nparr % 2 == 0)]

answered Sep 18, 2018 at 7:28

unlut

3,8122 gold badges16 silver badges24 bronze badges

Sign up to request clarification or add additional context in comments.

Comments

Deepak Saini · Accepted Answer · 2018-09-18 07:33:03Z

8

Since you don't provide an example data, use toy data:

# Cost of agents represented by indices of cost, we have agents 0, 1, 2, 3
cost = np.array([4,5,6,2])
# Agents to consider 
np_agents = np.array([0,1,3])
# threshold for each agent. Calculate different thresholds for different agents. Use array of indexes np_agents into cost array.
thresholds = cost[np_agents] # np.array([4,5,2])
feasible_agents = np_agents[np_agents > thresholds] # np.array([3])

answered Sep 18, 2018 at 7:33

Deepak Saini

2,9201 gold badge22 silver badges26 bronze badges

Collectives™ on Stack Overflow

Numpy filter using condition on each element

2 Answers 2

Comments

Comments

Your Answer

Linked

Hot Network Questions

Collectives™ on Stack Overflow

2 Answers 2

Comments

Comments

Your Answer

Sign up or log in

Post as a guest

Linked

Related