Manilos,
There is also scope of enhancing existing ML algorithms. In particular for Neural Net/ MLP adding more activation functions like Relu/ Tanh. Also adding functionality for deep learning architecture like CNN or LSTM which are gaining popularity. This may be more feasible in terms of limited timeframe and constrains pointed by Stephen. Get Outlook for Android From: Stephen Boesch Sent: Saturday, October 21, 8:39 AM Subject: Re: Add a machine learning algorithm to sparkml To: Manolis Gemeliaris Cc: dev@spark.apache.org A couple of less obvious facets of getting over the (significant!) hurdle to have an algorithm accepted into mllib (/spark.ml): the review time can be very long - a few to many months is a typical case even for relatively fast tracked algorithms you will likely be asked to provide evidence of a strong perceived need within the community/industry for the algorithm These considerations may make it challenging for you to find a yet-unimplemented algorithm that can be completed within a constrained timeframe. 2017-10-20 19:43 GMT-07:00 Manolis Gemeliaris <gemeliarismano...@gmail.com>: Hello everyone, I am an undergraduate student and now looking to do my final year project. Professor Minos Garofalakis suggested to me that as a project , I could find a machine learning algorithm not implemented by anyone ,in Spark.ml and implement it. As the topic is related to contributing code (an algorithm implementation) to Spark, I address to you also. My question to you is , are there any suggestions about what algorithm is missing from spark.ml currently that would be a good option to implement? (e.g. k-means and lda are already there and so is lsvm) Thanks in advance.