[sympy] Re: Proposal for Linear Algebra: Tensor core (Looking for the mentor)

2019-04-06 Thread zhiqikang
(Final version of proposal) Hello everyone, I have updated the final version of my proposal. Could anyone take a look at it and give me a feed back please? Here is the link: proposal Tensor core

[sympy] Re: Proposal for Linear Algebra: Tensor core (Looking for the mentor)

2019-04-06 Thread zhiqikang
(Final version of proposal) Hello everyone, I have updated the final version of my proposal. Could anyone take a look at it and give me a feed back please? Thank you in advance! Regards, Zhiqi -- You received this message because you are subscribed to the Google Groups "sympy" group. To

Re: [sympy] Re: Proposal for Linear Algebra: Tensor core (Looking for the mentor)

2019-04-02 Thread zhiqikang
Hello Aaron, First of all, thank you for this clarification! It makes the terms much more understandable. But there is one thing that seems confusing to me. Since NumPy is a numeric library and SymPy is a symbolic one, what is the point to have a NumPy-like array in this case? Does it mean the

Re: [sympy] Re: Proposal for Linear Algebra: Tensor core (Looking for the mentor)

2019-04-01 Thread Aaron Meurer
On Mon, Apr 1, 2019 at 5:26 PM wrote: > > Hello Aaron, > > I just updated my proposal by modifying what I can understand for now. Things > that I have updated are: > 1. Modifying and making sure the new operators for arrays are for symbolic > mathematics > 2. Adding a second part of this task

Re: [sympy] Re: Proposal for Linear Algebra: Tensor core (Looking for the mentor)

2019-04-01 Thread zhiqikang
Hello Aaron, I just updated my proposal by modifying what I can understand for now. Things that I have updated are: 1. Modifying and making sure the new operators for arrays are for symbolic mathematics 2. Adding a second part of this task by refering to an issue #15464

Re: [sympy] Re: Proposal for Linear Algebra: Tensor core (Looking for the mentor)

2019-04-01 Thread zhiqikang
Hello Aaron, Thank you for you replies. I think a have made a terrible mistake while preparing fot this project. In fact it was the sentence in the description of project idea > There are two branches: indexed symbols (which do not contain components > data), and array-like objects (containing

Re: [sympy] Re: Proposal for Linear Algebra: Tensor core (Looking for the mentor)

2019-04-01 Thread Aaron Meurer
One thing that worries me is that we shouldn't be reimplementing NumPy or other numeric libraries inside of SymPy. So something should be put in SymPy if it has a use for symbolic mathematics. Some things in your application it isn't clear if they fall into this category or not. I would try to

Re: [sympy] Re: Proposal for Linear Algebra: Tensor core (Looking for the mentor)

2019-03-31 Thread zhiqikang
Hello Aaron, I have properly updated my proposal in this link https://docs.google.com/document/d/1w_-hWrrT89Dr70Xo1TPkQdU80UIVku0Y9cNHADgdfJg/edit?usp=sharing Could you please review it and give me some advice if you are availble? I would sincerely appreciate it. I will continue to update

Re: [sympy] Re: Proposal for Linear Algebra: Tensor core (Looking for the mentor)

2019-03-29 Thread zhiqikang
Hello, Ok, I get it! So this lazy operation is an extension of the precedent tasks which are implementation of numpy like operators. As you say, theses operators should be added inside the array.I would update the proposal asap. Besides, I would like to discuss with you about the integration

Re: [sympy] Re: Proposal for Linear Algebra: Tensor core (Looking for the mentor)

2019-03-28 Thread Aaron Meurer
I think the way to do this is to have operations on array objects that don't evaluate by default. This would be very similar to the matrix expressions code, where something like MatMul or MatAdd doesn't evaluate, even on explicit matrices, unless you call doit() on it (doit() is what SymPy

Re: [sympy] Re: Proposal for Linear Algebra: Tensor core (Looking for the mentor)

2019-03-28 Thread zhiqikang
Hello Aaron, I found the Indexed class that you talked about. In my opinion, there is a difference between Indexed and Lazyarray. The Indexed is an abstraction of mathematical object with indices, that is a model to represent the tensor in general; while the lazy array is to perform a lazy

Re: [sympy] Re: Proposal for Linear Algebra: Tensor core (Looking for the mentor)

2019-03-28 Thread zhiqikang
Hello Aaron, I am thinking about something like this: class Lazyarray(object): > > """ > > Lazy-evaluated array > > """ > >def __init__(self, array): > > """ > > Initiate a Lazyarray with an array > > """ > > self.operations = [] > > def evaluate(self, f=None): > > """ > > Evaluate all

Re: [sympy] Re: Proposal for Linear Algebra: Tensor core (Looking for the mentor)

2019-03-28 Thread Aaron Meurer
Can you clarify what a lazy array would look like? I think this may already be implemented as Indexed. Aaron Meurer On Thu, Mar 28, 2019 at 11:20 AM wrote: > (updated: Lazy operators) > > Ø lazy operators on arrays > > lazy evaluation can improve the performance while iterating the

[sympy] Re: Proposal for Linear Algebra: Tensor core (Looking for the mentor)

2019-03-28 Thread zhiqikang
(updated: Lazy operators) Ø lazy operators on arrays lazy evaluation can improve the performance while iterating the array since it creates value only if it is called. To implement lazy operators, I am thinking about two plans: 1. Create a new sub-module named lazy-array