Mark Maunder

Check out my research at populationassessment.com

"It is with great disappointment that we must announce the cancellation of the 75th Tuna Conference at the UCLA Lake Arrowhead Lodge ... Due to unforeseen circumstances beyond our control ... we are unable to meet the minimum attendance required ..." Coordinated by SWFSC-IATTC

Hi Everyone, CAPAM’s special issue on Fisheries Stock Assessment Good Practices is finally completed. Direct link to the special issue is here sciencedirect.com/special-issue/… The preface that briefly describes the papers can be found at sciencedirect.com/science/articl…

@TrevorABranch But who is following you? I just joined and don't see much #fisheries action over there. But I do agree that X has nearly gone extinct in terms of fisheries.

My bluesky account has gone from 800 to 1800 followers in the last 3-4 weeks, despite minimal posts over that time. Definitely the place to be now. Meanwhile on Twitter I don't get new followers any more.

New paper on relating the #stockrecruitment relationship with evaluating the #stockstatus with respect to #referencepoints in #FisheriesManagement focusing on the depletion estimation and the double jeopardy of the SR relationship. mdpi.com/2410-3888/9/11…

@calcofi @Scripps_Ocean @noaafisheries @noaaocean @sanctuaries

How do we address rapid changes in fleet structure within integrated stock assessments? When is it necessary to represent sex-structured dynamics? Matt Cheng used Alaska sablefish and simulations to address these questions. See his @UAFcfos PhD defense on Monday 10/21 @ 1pm AK

Paper by John Hoenig on estimating #naturalmortality in the #CAPAM #GoodPractices special issue in #FisheriesResearch. Highlight need for contrast in fishing mortality and outline how this can be achieved over time, within a year, age or size, space. sciencedirect.com/science/articl…

New paper on #diagnostics by Maia Kapur in the #CAPAM #GoodPractices special issue of #FisheriesResearch Results further support using multiple diagnostics  sciencedirect.com/science/articl…

The top 4 most cited articles in the Journal #FisheriesResearch published since January 2021 are from #CAPAM special issues sciencedirect.com/journal/fisher…

It is part of the #CAPAM #GoodPractices Special Issue. Which is close to being complete with only two papers and the preface to come sciencedirect.com/special-issue/…

We have just published a paper describing the history of #CAPAM focusing on the workshop series and consequent special issues in #FisheriesResearch. #StockAssessment #model #fisheries #tuna authors.elsevier.com/a/1johtbiU21AhL

Some other good papers in the special issue sciencedirect.com/special-issue/…

Just published a paper on a framework for #populationdynamics #models in #EcologicalModelling part of a joint special issue on good #modelling practices. Based on #Fisheries #StockAssessment with references to #CAPAM work. Inc #Diagnostic #ExpertSystem authors.elsevier.com/sd/article/S03…

Fishery Officer (Fisheries Assessment and Management Advice) job at FAO jobs.fao.org/careersection/…

IATTC researchers and Kindai University are studying how rising ocean temperatures affect tuna larvae survival. Investigadores de la CIAT y la Universidad de Kindai estudian cómo el aumento de la temperatura oceánica afecta la supervivencia de las larvas de atún.

Also see medium.com/thirdai-blog/g…

Kind of off topic, but I wonder how much resources (CPUs, time, money) it would cost to train a LLM on wikipedia and your own data (e.g. all the pdfs of the CAPAM special issues)

I Was thinking more like Bloomburg-gpt bloomberg.com/company/press/…

Here are some costs @gilinachum/llm-domain-adaptation-using-continued-pre-training-part-1-3-e3d10fcfdae1" target="_blank" rel="nofollow noopener">medium.com/@gilinachum/ll…

@ErnestoCarrella @CubillosS_L "Continued Pre-training" Provide unlabeled data and Pre-training process will tweak the model parameters to accommodate the input data and improve its domain knowledge. Is the closest without complete training, but i assume will be limited. docs.aws.amazon.com/bedrock/latest…

@maunderCAPAM @CubillosS_L You can look up 'fine tuning' for GPT which is not massively expensive, but it sounds like what you want is indexing a larger number of documents and then retrieve the right page at the right time which probably can be done cheaper still

@CubillosS_L Creating prompts (RAG) from your pdfs or fine tuning the GPT is sub-optimal. Better to completely train the model with the data you want. Add your data to some general reliable data, without any extra fluff. Is that possible and how much resources would it need?

@maunderCAPAM Inestead of a LLM, may be implementing a GPT is enough (with a link to where pdfs are).