Can I use hadoop to train a neutral network?

Question

I want to train a neural network using Hadoop. We know that when training a neural network, the weights of each neuron change at each iteration, and each iteration depends on the previous one. I am new to Hadoop and not quite familiar with the features it provides. Is it possible to associate an iteration with the addDependingJob() method, emphasizing the dependencies? Or you can use other tricks to implement NN using Hadoop.

Any advice would be highly appreciated.

Thanks and best regards.

Answer 1

You can write it yourself. If you know how to write Bacvk propogation in one core from scratch. It can be easily ported to the Mapreduce approach. The HDFS cache should store the current weights of the neurons, and each map task should evaluate their update by the training instance, and then the reduction should summarize all these updates and put them in a cachce.