sandeepeecs

in india they are doing some thing called ONDC https://en.wikipedia.org/wiki/Open_Network_for_Digital_Comme...

ONDC is not an application, an intermediary, or software, but a set of specifications designed to foster open interchange and connections between shoppers, technology platforms, and retailers.[3] Technological self reliance, demand for level playing field mainly from small retailers, lower the barrier of entry and discovery online, adoption of open digital ecosystem across key sectors and fixing the non-competitive behavior of big ecommerce firms like Amazon and Flipkart to capture the US$810 billion domestic retail market led to its creation.[4] Designed to keep check on Big Tech companies from violating Consumer Protection (E-Commerce) (Amendment) Rules, 2021 due to concentration of market power by integrating them into an open-source decentralised network where data portability will break data silos while data interoperability will allow innovation.

Stripe Reader—our first piece of hardware—is a mobile card reader that works with Terminal’s APIs & SDKs to enable developers to build their own, custom in-person payments experiences.

When you are dealing with payments data x paying to y and txn volumes of few months then they are very large graphs, companies such as npci or alipay deals with these kinds of data.

Some of the usecase to build such graph is to get node embeddeding for fraud prevention or link prediction etc.

We have an implementation of this type of algorithm at https://alpes.ai that powers the API. This can be an alternative to deep neural network solutions

The context here is we are based out of Hyderabad(india) we do not have easy access to top universities mathematicians where the fundamental research happens in math.

We are trying : https://www.thehansindia.com/posts/index/Young-Hans/2018-11-...

As far as I know, I think he might have sent it to journals and conferences. but the problem is for them to consider it as it is a very big problem in math they look for some initial validation from the community I may be wrong.

This is my attempt to understand how the hacker news community would validate it.

its there on arxiv. https://arxiv.org/pdf/1609.06971v4.pdf

Looks like this is responded by Raghavan and this is what I had seen on research gate comments

Dr. Kumar's response to the above: I certainly know that the lambda - sequence is fixed and unalterable because each lambda(n) is obtained by factorizing n into prime factors and then defining lambda(n)=+1 if there are even factors else, lambda(n) =-1 if n is a prime or has odd prime factors.

In the paper for very long sequences, the lambda-sequence is treated as one instance of a hypothetical random walk. If this analogy is true then the magnitude of L(N), which is the sum of the first N terms of lambda(n)’s, (where N is very large) can be likened to the expected distance travelled by a random walker in N steps which is given by C .N^(1/2) (see S. Chandrasekhar(1943)).

However, for this analogy to be really meaningful and accurate, one must prove the lambda(n), for large and arbitrary n, must satisfy the criteria: (i) equal probabilities of being +1 or -1 , (ii) the lambda-sequence has no cycle and (iii) unpredictability.

In the paper I provide mathematical proofs for all the above criteria, after which one can deduce the asymptotic expression for L(N) as C. N^(1/2+e). We then invoke (i) Littlewoods Theorem 1 (proved in the paper) and then (ii) use Khinchin (1924) and Kolmogorov’s (1929) law of the iterated logarithm, for evaluating the bound ‘e’ and to show that e tends to zero as N tends to infinity, thus finally proving R.H.

One last comment: Herrington quotes Borwein’s statement as an “Equivalence to RH”, in actuality the condition stated by Borwein (2008) is only a necessary condition for RH to be true. The additional criteria (above) needs to be satisfied and hence need to be proved as done in my paper.

Yes I agree. most people who come to us at alpes AI do care about training time. how fast they can do experiments

Another important aspect is training and incremental training on edge device.

At the time when privacy is becoming very important and you cannot export data from mobile devices etc. Training time on mobile is an important factor

Its great to see some new way of looking at things in AI.

We are also attempting to developing an algorithm which learns without doing a backpropagation more details are http://alpes.ai

We are an AI company working out of Hyderabad / India, We have a very fast Classification algorithm as a product. and currently been used by 3 paying customers.

http://www.alpes.ai/

We are releasing our API next week. The idea being you can quickly try out the algorithm and see Results and Training time on your own datasets.

hey paper is on its way probably we will be publishing it in couple of weeks from now.

Nice to see great advancements in the training speed for deep neural network algorithms. We are also very interested in improving the speed of training. We at alpes.ai have developed a non-recursive neural network algorithm which has a very fast training time. We got very good results for standard open datasets, training time in the range of seconds and accuracy on par with standard results on normal laptops without any special GPU`s or hardware.

http://alpes.ai/

These are the results of some of the datasets

DataSet Training Time Accuracy

Extended Yale DB 40 sec 94.00%

Human Activity Detection Dataset 3 sec 86%

MNIST 90 sec 97%

Google Speech Dataset 60 sec 92%

Liver Dataset 2 sec 89%

actually its quite usefull project. Do you know. we can use this for campaign like we love Iran. http://www.weloveiran.org/

can you share the techonology stack or how its built. is it open sourced ?