3 ms·
The funny thing is that the authors of the paper he linked actually answer his question in the first paragraph, when they say that the input dataset needs to be
by low_tech_love 3y ago
The funny thing is that the authors of the paper he linked actually answer his question in the first paragraph, when they say that the input dataset needs to be significantly larger than the number of weights to achieve good generalisation, but there is usually not enough data available.