How to use pre-trained models in your next business project

Most of the new deep learning models being released, especially in NLP, are very, very large: They have parameters ranging from hundreds of millions to tens of billions.

Given good enough architecture, the larger the model, the more learning capacity it has. Thus, these new models have huge learning capacity and are trained on very, very large datasets.

Because of that, they learn the entire distribution of the datasets they are trained on. One can say that they encode compressed knowledge of these datasets. This allows these models to be used for very interesting applications—the most common one being transfer learning. Transfer learning is fine-tuning pre-trained models on custom datasets/tasks, which requires far less data, and models converge very quickly compared to training from scratch.

Read: [How machines see: everything you need to know about computer vision]

How pre-trained models are

Read More

Wells Fargo to donate $400 million in PPP fees to small business recovery effort

Wells Fargo detailed plans Thursday to donate $400 million in processing fees it earned from the Paycheck Protection Program to an initiative it says will help small businesses recover from the impacts of COVID-19.

In April, NPR reported that banks took in over $10 billion in fees from the federal government for handling the Paycheck Protection Program, which is administered by the Small Business Administration and designed to provide relief for small businesses.

The bank announced its recovery effort, the “Open for Business” fund, in a news release Thursday morning.

The fund will work with nonprofits that provide capital, technical support and programs for resiliency, for example by helping a business pivot its supply chain, said Jenny Flores, head of Small Business Growth Philanthropy for the bank.

The first round of funding will give $28 million to community lenders known as Community Development Financial Institutions that help Black-owned small businesses,

Read More

Bank of Canada Says Business Sentiment Is at Lowest Since 2009

(Bloomberg) — Canadian business sentiment has fallen to its lowest level since the 2008-2009 recession as sales slow and uncertainty about future growth remains elevated, according to a survey of executives released Monday by the Bank of Canada.

The Ottawa-based central bank polled businesses between May 12 and June 5 to gauge sentiment during the pandemic. The results show that even as provinces begin to reopen their economies, many businesses are still struggling with weak demand.

The plunge in sentiment is hardly a surprise, given the nation fell into its deepest recession last quarter since the Great Depression. While there are some positive notes with the central bank highlighting that many businesses expect a fairly quick rebound, the overall gist of the data paints a business sector that has suffered a major shock.

Results “suggest that business sentiment is strongly negative in all regions and sectors due to impacts from

Read More

Amazon’s business practices examined by two U.S. states

(Reuters) – State investigators in both California and Washington are examining Amazon’s business practices, two newspapers reported, citing people familiar with the matter.

The inquiries focus at least in part on how Amazon treats sellers in its online marketplace, the Wall Street Journal and the New York Times reported on Friday.

Amazon declined to comment on the report of an investigation by California. It was not immediately available for comment outside regular market hours on the Washington review report.

The reports said California state is reviewing Amazon’s practices on selling its own products in competition with third-party vendors. Amazon said its in-house products only account for about 1% of its total annual retail sales.

Washington state is also investigating whether Amazon makes it harder for sellers to list their products on other websites, the NYT reported. The inquiries did not appear to be in advanced stages, the NYT report added,

Read More